Psychosocial stress and cardiovascular disease: The ... - UCL Discovery [PDF]

acute intravenous pre-treatment with propranolol (Jezová, Vigas, Klimes, & Jurcovicová,. 1983). In agreement with

8 downloads 3 Views 3MB Size

Recommend Stories


Untitled - UCL Discovery
Be who you needed when you were younger. Anonymous

Nutrition and Cardiovascular Disease
Suffering is a gift. In it is hidden mercy. Rumi

Alcohol and Cardiovascular Disease
Don’t grieve. Anything you lose comes round in another form. Rumi

Homocysteine and cardiovascular disease
We may have all come on different ships, but we're in the same boat now. M.L.King

[PDF] Diabetes in Cardiovascular Disease
Never let your sense of morals prevent you from doing what is right. Isaac Asimov

HDL and cardiovascular disease
You miss 100% of the shots you don’t take. Wayne Gretzky

icmr psychosocial stress scale
Before you speak, let your words pass through three gates: Is it true? Is it necessary? Is it kind?

cardiovascular disease
This being human is a guest house. Every morning is a new arrival. A joy, a depression, a meanness,

Cardiovascular Disease
This being human is a guest house. Every morning is a new arrival. A joy, a depression, a meanness,

Cardiovascular Disease
Knock, And He'll open the door. Vanish, And He'll make you shine like the sun. Fall, And He'll raise

Idea Transcript


Psychosocial stress and cardiovascular disease: The role of the hypothalamic-pituitary-adrenal axis

Amy Ronaldson 2016 University College London

Thesis submitted for the degree of Doctor of Philosophy

1

Student Declaration I, Amy Ronaldson, confirm that the work presented in this thesis is my own. Where information has been derived by other sources, I confirm that this has been indicated in the thesis.

Signed: __________________

Date: _____________

2

Acknowledgements Firstly I would like to thank my supervisor Professor Andrew Steptoe for helping me in obtaining the funding for this studentship, and providing invaluable guidance and support throughout. Furthermore, I would like to thank him for giving me the opportunity to work within the Psychobiology Group in the first place. If I make it in academia I will have Professor Steptoe to thank! I would also like to thank Dr Livia Carvalho for her help and guidance in the laboratory. Without Dr Carvalho’s help the Stress Pathways Study would not have been possible. I would also like to thank the British Heart Foundation for funding my studentship. I would have really struggled these past three years without my wonderful colleagues here in the Psychobiology Group. During this time many colleagues have become dear friends and I would not have made it to the end of my PhD studies without the support, advice, and comic relief provided by my office mates Lydia Poole, Stephanie Schrempft, Sam Lawes, and fellow Irishwoman Ruth Hackett. They have spent the last three years listening to my moaning and complaining and apparently still like me. I also have to thank my housemates (and best friends) for putting up with months of mood swings and complaining. Nic and Keavy – you are amazing. Importantly, I need to thank my special canine housemate Ivy for giving me lots of much needed cuddles over the last year. Last but not least I have to thank my wonderful family for their help and support. Thank you Mum and Dad for never pushing me too hard and letting me discover my own way here.

3

Abstract There is evidence to suggest that dysregulation of the HPA axis might be one of the biological pathways linking psychosocial stress with cardiovascular disease (CVD). This PhD consisted of three studies that aimed to assess the role of HPA axis dysregulation in CVD, and to examine potential biological mechanisms that might be involved in stressrelated HPA axis dysregulation. Study 1 assessed the utility of pre-surgical diurnal cortisol rhythm in predicting adverse outcomes in advanced heart disease using data from an observational clinical cohort study. The results showed that patients with flatter pre-surgical cortisol slopes were at increased risk of experiencing an adverse event in the years following coronary revascularisation. This finding provides evidence for the clinical relevance of HPA axis dysregulation in CVD. Study 2 and 3 sought to garner more information about the biological mechanisms underlying stress-related HPA axis dysregulation using data from a randomised controlled trial involving the administration of pharmacological probes to healthy volunteers. In Study 2 the effects of six-day administration of beta-blockers and SSRIs on diurnal cortisol secretion were examined. Results indicated that women taking SSRIs had significantly steeper diurnal cortisol slopes compared to placebo. Mechanistically, these results support the notion that the serotonergic system exerts substantial effects on the HPA axis, potentially through modulation of the serotonergic or corticosteroid receptors. Therapeutically, these results suggest that SSRIs might be a plausible intervention for female CHD patients with flatter cortisol slopes

4

In Study 3 the effects of seven-day administration of beta-blockers and SSRIs on cortisol stress reactivity and corticosteroid receptor sensitivity in the laboratory were investigated. The results indicated that generally, acute stress brought about a decrease in corticosteroid receptor sensitivity. SSRIs enhanced this decrease and also blunted the cortisol stress response. These results suggest that SSRIs may enhance adaptive stress-related changes in HPA axis function, thereby having therapeutic implications for stress-related illness such as CVD. Together this body of work indicates that alterations in HPA axis function play a role in CVD and that the serotonergic system likely plays a role in stress-related dysregulation of the HPA axis.

5

Table of Contents ACKNOWLEDGEMENTS ............................................................................................ 3 ABSTRACT ..................................................................................................................... 4

CHAPTER 1

LITERATURE REVIEW: STRESS AND CARDIOVASCULAR DISEASE.............................................................................................. 18

1.1 Introduction ................................................................................................................ 18 1.2 Cardiovascular disease: Pathogenesis and prevalence ..................................... 18 1.3 Stress and cardiovascular disease: Introduction ............................................... 22 1.4 Stress and cardiovascular disease: External stressors ..................................... 23 1.4.1

External stressors: Perceived stress and chronic stress burden............... 23

1.4.2

External stressors: Caregiver stress ............................................................. 27

1.4.3

External stressors: Work stress, financial stress, and social isolation .... 28

1.4.4

External stressors: Summary ......................................................................... 31

1.5 Stress and cardiovascular disease: Negative emotional disorders ................. 32 1.5.1

Negative emotional disorders: Psychological distress ............................. 32

1.5.2

Negative emotional disorders: Depression and anxiety ............................ 33

1.5.3

Negative emotional disorders: Summary ..................................................... 35

1.6 Stress and cardiovascular disease: Acute stress triggers ................................. 36 1.7 Stress and cardiovascular disease: Prognosis in those already affected ...... 40 1.7.1

The role of external stressors in CVD prognosis ........................................ 41

1.7.2

The role of negative emotional disorders in CVD prognosis .................... 43

1.8 Chapter summary ...................................................................................................... 46 CHAPTER 2

LITERATURE REVIEW: THE ROLE OF THE HYPOTHALAMIC-PITUITARY-ADRENAL AXIS AND THE CORTICOSTEROID RECEPTORS ................................................ 49

2.1 Introduction ................................................................................................................. 49 2.2 Potential pathways linking psychosocial stress and CVD ............................... 49 2.3 The stress system and the stress response ............................................................ 50 2.4 Stress-related HPA axis activity ............................................................................. 53 2.5 Diurnal HPA axis activity ........................................................................................ 55 2.6 Psychosocial stress and dysregulation of diurnal HPA axis function ............ 56 6

2.6.1

Psychosocial stress and the CAR .................................................................. 56

2.6.2

Psychosocial stress and the cortisol slope ................................................... 59

2.6.3

Psychosocial stress and cortisol AUC ......................................................... 61

2.7 Psychosocial stress and dysregulation of stress-related HPA axis activity ... 63 2.7.1

Exposure to chronic stress and early life adversity .................................... 63

2.7.2

The effects of depression ............................................................................... 65

2.8 HPA axis dysregulation and CVD .......................................................................... 67 2.8.1

Diurnal HPA axis function in CVD .............................................................. 68

2.8.2

Cortisol stress reactivity in CVD .................................................................. 70

2.9 The role of the corticosteroid receptors ............................................................... 72 2.9.1

What causes modulation of corticosteroid receptor sensitivity? .............. 74

2.9.2

How do we measure corticosteroid receptor sensitivity? .......................... 77

2.10 Psychosocial stress and the corticosteroid receptors ........................................ 80 2.10.1 Chronic stress and corticosteroid receptor sensitivity ............................... 80 2.10.2 Acute stress and corticosteroid receptor sensitivity ................................... 85 2.11 Chapter summary ...................................................................................................... 92

CHAPTER 3 STUDY 1 – DIURNAL CORTISOL RHYTHM AND ADVERSE CLINICAL OUTCOMES IN PATIENTS WITH ADVANCED CVD: THE ARCS STUDY ................................................................... 94 3.1 The Adjustment and Recovery after Cardiac Surgery (ARCS) Study ........ 94 3.2 My contribution to the ARCS Study .................................................................... 95 3.3 Differentiating my PhD from the ARCS Study ................................................... 96 3.4 Introduction ................................................................................................................. 96 3.4.1

Hypotheses ....................................................................................................... 97

3.5 Materials and methods ............................................................................................. 98 3.5.1

Participants ...................................................................................................... 98

3.5.2

Biological and clinical measures .................................................................. 99

3.5.3

Psychosocial stress variables ...................................................................... 101

3.5.4

Covariates: clinical and sociodemographic factors ................................. 104

3.5.5

Statistical analyses ........................................................................................ 105

3.6 Results ........................................................................................................................ 106 3.6.1

Pre-surgical cortisol and clinical outcomes ............................................. 108 7

3.6.2

Pre-surgical cortisol slope and psychosocial stress variables ............... 112

3.7 Discussion .................................................................................................................. 113 3.7.1

Summary of results ....................................................................................... 113

3.7.2

Comparison to previous research .............................................................. 114

3.7.3

Potential mechanisms explaining the link between diurnal cortisol rhythm and adverse outcomes ..................................................................... 115

3.7.4

CAR: Lack of association ............................................................................. 116

3.7.5

Psychosocial stress: Lack of association ................................................... 117

3.7.6

Strengths and limitations.............................................................................. 118

3.7.7

Conclusion...................................................................................................... 121

CHAPTER 4 THE STRESS PATHWAYS STUDY: INTRODUCTION AND METHODS ......................................................................................... 122 4.1 The Stress Pathways Study .................................................................................... 122 4.2 Differentiating my PhD from the Stress Pathways Study .............................. 123 4.3 The pharmacological probes.................................................................................. 124 4.3.1

Beta blocker: Propranolol .......................................................................... 125

4.3.2

SSRI: Escitalopram ...................................................................................... 127

4.4 Method ........................................................................................................................ 129 4.4.1

Study design .................................................................................................. 129

4.4.2

Sample size .................................................................................................... 129

4.4.3

Participants ................................................................................................... 130

4.4.4

Study protocol ............................................................................................... 131

4.4.5

Psychosocial measures ................................................................................ 136

4.4.6

Adverse events and drug effects ................................................................. 141

4.4.7

Biological measures ..................................................................................... 141

4.5 Data storage ............................................................................................................... 144 4.6 Statistical analyses ................................................................................................... 144 4.7 My contribution to the Stress Pathways Study ................................................. 145

CHAPTER 5 STUDY 2 – THE STRESS PATHWAYS STUDY RESULTS: THE EFFECT OF PHARMACOLOGICAL BLOCKADE ON DIURNAL CORTISOL SECRETION IN HEALTHY VOLUNTEERS ......... 146 5.1 Introduction ............................................................................................................... 146 8

5.2 Literature review: Beta-blockers and basal HPA axis function.................... 146 5.2.1

Acute administration of beta-blockers........................................................ 147

5.2.2

Long-term administration of beta-blockers .............................................. 148

5.3 Literature review: SSRIs and basal HPA axis function .................................. 150 5.3.1

Acute administration of SSRIs ..................................................................... 150

5.3.2

Long-term administration of SSRIs............................................................. 152

5.3.3

SSRIs and diurnal HPA axis function ......................................................... 155

5.4 Aims and hypotheses ............................................................................................... 157 5.5 Calculation of diurnal cortisol parameters ........................................................ 158 5.6 Statistical analyses ................................................................................................... 160 5.7 Results ......................................................................................................................... 161 5.7.1

Participants ................................................................................................... 161

5.7.2

Study medication effects on stress-related psychological factors ......... 162

5.7.3

Study medication effects on diurnal cortisol parameters ........................ 164

5.8 Discussion ................................................................................................................... 170 5.8.1

Summary of results ....................................................................................... 170

5.8.2

Comparison to previous research .............................................................. 170

5.8.3

Potential mechanisms explaining SSRI effects on the HPA axis ............ 172

5.8.4

Sex differences in SSRI effects and HPA axis function ............................ 173

5.8.5

Therapeutic implications .............................................................................. 175

5.8.6

Beta-blockers: Lack of effect ....................................................................... 176

5.8.7

Strengths and limitations.............................................................................. 177

5.8.8

Conclusion...................................................................................................... 179

CHAPTER 6 STUDY 3 – THE STRESS PATHWAYS STUDY RESULTS: THE EFFECT OF PHARMACOLOGICAL BLOCKADE ON CORTISOL STRESS REACTIVITY AND CORTICOSTEROID RECEPTOR SENSITIVITY IN HEALTHY VOLUNTEERS ........180 6.1 Introduction .............................................................................................................. 180 6.2 Literature review: Beta-blockers and cortisol stress reactivity ................... 180 6.2.1

Acute administration of beta-blockers........................................................ 181

6.2.2

Long-term administration of beta-blockers .............................................. 183

6.2.3

Summary ......................................................................................................... 184

6.3 Literature review: SSRIs and cortisol stress reactivity ................................... 185 9

6.3.1

Acute administration of SSRIs ..................................................................... 185

6.3.2

Long-term administration of SSRIs............................................................. 186

6.3.3

Summary ......................................................................................................... 187

6.4 Cortisol stress reactivity: Aims and hypotheses................................................ 188 6.5 Literature review: Beta-blockers and corticosteroid receptor sensitivity .. 189 6.6 Literature review: SSRIs and corticosteroid receptor sensitivity................. 190 6.6.1

Murine studies ............................................................................................... 190

6.6.2

Human in vitro studies.................................................................................. 191

6.6.3

Human in vivo studies .................................................................................. 192

6.6.4

Summary ......................................................................................................... 195

6.7 Corticosteroid receptor sensitivity: Aims and hypotheses ............................. 196 6.8 Biological measures.................................................................................................. 198 6.8.1

Calculation of cortisol parameters in the laboratory............................... 199

6.8.2

Corticosteroid receptor sensitivity .............................................................. 200

6.8.3

Cardiovascular measures............................................................................. 202

6.9 Statistical analyses ................................................................................................... 203 6.10 Results ......................................................................................................................... 205 6.10.1 Participants .................................................................................................... 205 6.10.2 Subjective stress ratings ............................................................................... 207 6.10.3 Cardiovascular measures............................................................................. 210 6.10.4 Cortisol stress reactivity............................................................................... 217 6.10.5 Corticosteroid receptor sensitivity .............................................................. 221 6.10.6 Sensitivity analyses ....................................................................................... 228 6.11 Discussion ................................................................................................................... 229 6.11.1 Aims and hypotheses ..................................................................................... 229 6.11.2 Summary of results ........................................................................................ 229 6.11.3 Stress-related changes in cardiovascular measures ................................ 231 6.11.4 Cortisol stress reactivity............................................................................... 232 6.11.5 Stress-related changes in corticosteroid receptor sensitivity in healthy unmedicated volunteers ................................................................................ 233 6.11.6 Stress-related changes in corticosteroid receptor sensitivity in those receiving propranolol ................................................................................... 235 6.11.7 Stress-related changes in corticosteroid receptor sensitivity in those receiving escitalopram ................................................................................. 236 10

6.11.8 Cortisol stress reactivity: Lack of response............................................... 239 6.11.9 Strengths and limitations.............................................................................. 241 6.11.10 Conclusion .................................................................................................... 245

CHAPTER 7 DISCUSSION ..................................................................................... 246 7.1 Overview .......................................................................................................... 246 7.2 Main findings ................................................................................................... 247 7.2.1

Study 1: Diurnal cortisol rhythm and adverse clinical outcomes in patients with advanced CVD: The ARCS Study ................................... 247

7.2.2

Study 2: The effect of pharmacological blockade on diurnal cortisol secretion in healthy volunteers ............................................................. 250

7.2.3

Study 3: The effect of pharmacological blockade on cortisol stress reactivity and corticosteroid receptor sensitivity in healthy volunteers ............................................................................................................... 252

7.3 Overall summary of findings and implications ............................................ 256 7.3.1

Study 1: Implications............................................................................. 256

7.3.2

Study 2: Implications............................................................................. 259

7.3.3

Study 3: Implications............................................................................. 261

7.4 Methodological issues and limitations ........................................................... 263 7.4.1

The study samples ................................................................................. 263

7.4.2

The HPA axis: Measurement issues ...................................................... 266

7.4.3

The study medications ........................................................................... 268

7.5 Suggestions for future research ..................................................................... 269 7.5.1

The clinical utility of HPA axis dysregulation in CVD ......................... 269

7.5.2

Alternative pharmacological probes ..................................................... 270

7.5.3

The role of arginine vasopressin ........................................................... 273

7.6 Final conclusion ............................................................................................... 275

References .................................................................................................................... 277 List of publications ...................................................................................................... 330 Conference presentations ........................................................................................... 333 Appendices ................................................................................................................... 334

11

List of Tables

Table 2.1

Studies examining the effects of chronic stress on corticosteroid receptor function ................................................................................................... 82

Table 2.2

Studies examining the effects of acute stress on corticosteroid receptor function ................................................................................................... 89

Table 3.1

Demographic, cortisol, and clinical characteristics of the sample .........106

Table 3.2

Results of Cox regression analysis (cortisol slope) ..............................109

Table 3.3

Results of Cox regression analysis (cortisol AUC) ...............................111

Table 3.4

Results of Cox regression analysis (CAR) .............................................111

Table 3.5

Cross-sectional associations between pre-surgical cortisol slope, and waking and evening levels, and psychosocial stress variables ..............112

Table 5.1

Demographic characteristics of the sample............................................163

Table 5.2

Mean cortisol parameter values and p values from ANOVAs comparing the effects of escitalopram and propranolol to placebo .........................166

Table 6.1

Demographic characteristics of the sample............................................208

Table 6.2

Subjective stress ratings at baseline, post-stress, and 20 minutes after stress .......................................................................................................211

Table 6.3

Mean values on cardiovascular measures in the laboratory in each experimental condition ...........................................................................213

Table 6.4

Mean (raw) cortisol values across the laboratory session in each experimental condition (p values from analyses with log transformed cortisol values) ......................................................................................218

Table 6.5

Raw IL-6 values (LPS) and mean log IC50 values (dexamethasone and prednisolone) across the laboratory session in each experimental condition.................................................................................................224

12

List of Figures Figure 1.1

The stages of atherosclerosis ................................................................... 20

Figure 2.1

The hypothalamic-pituitary-adrenal axis ................................................ 52

Figure 2.2

Four types of allostatic load .................................................................... 54

Figure 2.3

The diurnal nature of HPA axis function ................................................ 55

Figure 2.4

Translocation of the corticosteroid receptor into the cell nucleus .......... 74

Figure 3.1

Flow diagram of participant recruitment and attrition from the ARCS Study ......................................................................................................102

Figure 3.2

Mean salivary cortisol values (ARCS Study) ........................................109

Figure 3.3

Kaplan-Meier survival curves for patients split into two equal groups at the median diurnal cortisol slope ...........................................................110

Figure 4.1

Flow diagram of participant data collection and attrition from the Stress Pathways Study .....................................................................................132

Figure 4.2

Study protocol: Blood and saliva sampling ...........................................134

Figure 4.3

Flow diagram of the stress protocol with subjective stress ratings ........136

Figure 5.1

Propranolol versus placebo: Mean salivary cortisol values across the day averaged across men and women ...........................................................165

Figure 5.2

Sex differences in the CAR across groups in both sets of pairwise analyses ..................................................................................................165

Figure 5.3

The interactive effect of sex on cortisol slope (wake-bedtime difference) within participants receiving escitalopram.............................................167

Figure 5.4

Escitalopram versus placebo: Mean salivary cortisol values across the day in men and women ..........................................................................169

Figure 6.1

The glucocorticoid sensitivity assay ......................................................201

Figure 6.2

Sample calculation of the IC50 from one participant ..............................203

Figure 6.3

Heart rate, SBP, DBP, and cardiac index values across the testing session in the overall sample ..............................................................................212

Figure 6.4

Mean heart rate values across the session in the propranolol and placebo groups. ....................................................................................................215

Figure 6.5

Mean cardiac index values across the session in the escitalopram and placebo groups. ......................................................................................216

Figure 6.6

Mean cortisol values (not log-transformed) at each time-point across the testing session in the overall sample ......................................................217

Figure 6.7

Mean cortisol values (not log-transformed) at each time-point across the testing session in the propranolol, escitalopram, and placebo groups. ..220

Figure 6.8

Mean log IC50 values for dexamethasone in the placebo group.............222 13

Figure 6.9

Mean log IC50 values for prednisolone in the placebo group.................223

Figure 6.10

Mean log IC50 values for dexamethasone in the propranolol and placebo groups .....................................................................................................225

Figure 6.11

Mean log IC50 values for prednisolone in the propranolol and placebo groups ....................................................................................................225

Figure 6.12

Mean log IC50 values for dexamethasone in the escitalopram and placebo groups .....................................................................................................227

Figure 6.13

Mean log IC50 values for prednisolone in the escitalopram and placebo groups .....................................................................................................227

14

List of Abbreviations

11β-HSD

11β-hydroxysteroid dehydrogenase

5-HTP

5-hydroxytryptophan

ACE

Angiotensin converting enzyme

ACS

Acute coronary syndrome

ACTH

Adrenocorticotropic hormone

AP-1

Activator protein-1

ARCS

Adjustment and recovery after cardiac surgery

AUC

Area under the curve

AVP

Arginine vasopressin

BDI

Beck Depression Inventory

BMI

Body mass index

BP

Blood pressure

CABG

Coronary artery bypass graft

CAR

Cortisol awakening response

CBG

Corticosteroid-binding globulin

CI

Confidence interval

CHD

Coronary heart disease

CNS

Central nervous system

CRH

Corticotropin releasing hormone

CRP

C-reactive protein

CV

Coefficient of variation

CVD

Cardiovascular disease

DBP

Diastolic blood pressure

DEX

Dexamethasone 15

DST

Dexamethasone suppression test

ERI

Effort-reward imbalance

ESSI

ENRICHD Social Support Instrument

EuroSCORE European System for Cardiac Operative Risk Evaluation GAD

Generalised anxiety disorder

GHQ

General Health Questionnaire

GLM

General linear model

GR

Glucocorticoid receptor

GRE

Glucocorticoid response element

HADS

Hospital Anxiety and Depression Scale

HPA

Hypothalamic-pituitary-adrenal

HSP-90

Heat shock protein 90

IC50

Inhibitory concentration 50%

IFN-γ

Interferon-γ

IL

Interleukin

IL-1Ra

Interleukin-1 receptor antagonist

LC-NE

Locus coeruleus-norepinephrine

LPS

Lipopolysaccharide

M

Molar

MACE

Major adverse cardiac event

MCP-1

Monocyte chemotactic protein 1

MI

Myocardial infarction

MR

Mineralocorticoid receptor

mRNA

Messenger RNA

NF-κB

Nuclear factor-kappa B

OC

Oral contraception

p23

Protein 23 16

PANAS

Positive and negative affect scale

PBMC

Peripheral blood mononuclear cells

PHA

Phytohaemagglutinin

PRED

Prednisolone

PSS

Perceived stress scale

PST

Prednisolone suppression test

PTSD

Post-traumatic stress disorder

RCT

Randomised controlled trial

SAM

Sympatho-adrenal-medullary

SBP

Systolic blood pressure

SES

Socioeconomic status

SNP

Single nucleotide polymorphism

SNS

Sympathetic nervous system

SRO

Social reorganisation stress

SNRI

Selective norepinephrine reuptake inhibitor

SSRI

Selective serotonin reuptake inhibitor

T2DM

Type 2 diabetes

TNF-α

Tumour necrosis factor-α

TSST

Trier Social Stress Test

WC

Waist circumference

WHO

World Health Organisation

17

Chapter 1 Literature review: Stress and cardiovascular disease 1.1 Introduction This chapter will describe the literature relating to the role of stress in cardiovascular disease (CVD). Firstly, the pathophysiology of CVD will be described. Following this, evidence for the role of psychosocial stress in the aetiology of CVD will be provided with a particular focus on external stressors, such as work stress, financial stress, and caregiver stress; negative emotional disorders, such as depression and anxiety; and acute stress triggers, such as natural disasters, war and terrorism, and periods of intense emotion. Additionally, this chapter will describe the literature on the effects of psychosocial stress on prognosis in those already diagnosed with CVD. The aim of this chapter is to highlight the importance of psychosocial stress in CVD progression and prognosis, while highlighting some of the limitations of the work to date. 1.2 Cardiovascular disease: Pathogenesis and prevalence CVD is an umbrella term referring to a group of diseases affecting the circulatory system. The most common forms of CVD are coronary heart disease (CHD) and stroke. Atherosclerosis is the primary pathological process underlying the development of CHD. It is a lifelong process whereby fatty deposits lead to the progressive narrowing of the coronary arteries due to the formation of atheromatous plaques. The lipid hypothesis of atherosclerosis holds that it is primarily a cholesterol storage disease. However, it is now understood that atherosclerosis is also an inflammatory disorder which can affect all middle- and large-sized blood vessels in the circulatory system (Hansson & Libby, 2006; Libby, Ridker, & Hansson, 2011). Atherosclerosis begins in childhood, with atherosclerotic change and development occurring during adolescence and young 18

adulthood (McGill et al., 2000; Williams et al., 2002). Across the lifespan, the cumulative effect of known cardiovascular risk factors accelerates the progression of atherosclerosis. These risk factors include clinical, biological, behavioural, and social factors. The clinical factors include hypertension, dyslipidaemia, type 2 diabetes, and overweight/obesity. Biological factors include genetic predisposition, older age, and being male. Behavioural factors include smoking, sedentary lifestyle, excessive alcohol intake, and poor diet. Low socioeconomic status (SES) and low education comprise the social factors. The majority of people who develop advanced atherosclerosis will be in an asymptomatic disease state for many years. The human artery contains three layers (See Figure 1.1, Box 1). The inner layer is called the tunica intima and is lined by a layer of endothelial cells. The next layer is the media, followed by the adventitia which contains nerve endings, microvessels, mast cells, and fibroblasts. Dyslipidaemia, hypertension, and the presence of pro-inflammatory cytokines can cause irritation to the endothelial cells lining the tunica intima. These endothelial cells then express adhesion molecules which capture leukocytes on their surface, when ordinarily white blood cells stream past without attaching. These leukocytes, which are primarily monocytes, then migrate into the intima where they mature into macrophages. The macrophages become resident in the artery wall and engulf lipoprotein molecules thus becoming foam cells (See Figure 1.1, Box 2). They also have a number of proinflammatory functions producing high levels of cytokines such as IL-1β and tumour necrosis factor. The development of atheromatous plaques also involves the migration of smooth muscle cells (the endogenous cells of the artery wall) from the media into the tunica intima where they proliferate forming a complex extracellular matrix through the release of macromolecules such as collagen and proteoglycans (See Figure 1.1, Box 3). This extracellular matrix forms a fibrous cap that covers the plaque. Underneath this cap, 19

the macrophages in the plaque begin to die via apoptosis thus releasing the lipids they have engulfed. The cellular debris and lipid molecules form a lipid-rich centre referred to as the necrotic core of the plaque. As cells and lipids accumulate the plaque enlarges and bulges into the lumen of the artery. Over time, the fibrous cap becomes thin and can fracture. If the plaque ruptures, the necrotic core of the plaque can leak into the lumen triggering the development of a thrombus (See Figure 1.1, Box 4).

1

2

3

4

Figure 1.1. The stages of atherosclerosis. Box 1 shows the cell structure of a healthy human artery. Boxes 2, 3, and 4 show the gradual progression of atherosclerosis culminating in plaque rupture. Adapted from Libby, Ridker, & Hansson (2011)

Plaques can cause clinical manifestations of CVD by either bringing about stenoses that limit blood flow to certain tissues leading to ischaemia, or by creating thrombi that lodge in arteries and interrupt blood flow. These clinical manifestations of CVD include acute coronary syndromes (ACS), namely myocardial infarction (MI) and unstable angina, as 20

well as stable angina. MI occurs when one of the coronary arteries is occluded by a thrombus following the rupture of an atheromatous plaque. The resulting ischaemia can lead to damage or death of cardiac tissue. Stable angina is a chronic condition characterised by chest pain on exertion caused by a lack of oxygen supply to the heart due to stenosis brought about by atherosclerosis. Unstable angina is distinct from stable angina in that chest pain occurs more frequently and for longer, and is not necessarily triggered by exertion. Unlike stable angina, unstable angina is caused by a thrombus partially occluding a coronary artery. Recent estimates from the World Health Organisation (WHO) revealed that CVD is the leading cause of death worldwide (WHO, 2015). In 2012 an estimated 17.5 million people died from CVD, accounting for 31% of all global deaths. Roughly 7.4 million of these deaths were caused by CHD and 6.7 million were due to stroke. Recent statistics from the UK have revealed that CVD is the second main cause of death after cancer with these diseases causing 27% and 29% of all deaths in 2014 respectively (Townsend, Bhatnagar, Wilkins, Wickramasinghe, & Rayner, 2015). In 2014, CVD accounted for around 155,000 deaths in the UK – approximately 69,000 deaths were due to CHD, and 39,000 were due to stroke (ibid). This makes CHD the biggest single cause of death in the UK accounting for 15% of male and 10% of female deaths (ibid). In the UK, CVD mortality rates have been in decline since the 1970s. Recent statistics from the British Heart Foundation show that between 1974 and 2013 CHD mortality rates have declined by 73% in those dying at any age, and 81% in those dying before 75 years (ibid). This reduction in mortality is thought to be attributable to a combination of reductions in major risk factors such as smoking, as well as improved hospital treatment and better clinical management of hypertension and dyslipidaemia (O’Flaherty, Buchan, & Capewell, 2013; Smolina, Wright, Rayner, & Goldacre, 2012). Despite the reduction in CVD mortality 21

rates, the economic costs of the disease are vast. In 2013/2014, the CVD healthcare expenditure within the UK amounted to approximately £5.9 billion (Townsend et al., 2015). Moreover, the total cost of CVD to the UK economy was estimated to be £15.2 billion in 2014 with this figure being attributable to direct healthcare costs, productivity losses, and informal care of CVD patients (ibid). A recent report by the Centre for Economics and Business Research predicts that the total costs of CVD to the UK will rise to £18.7 billion by 2020 (Centre for Economics and Business Research, 2014). 1.3 Stress and cardiovascular disease: Introduction As mentioned in the previous section, there are a number of well-established clinical, biological, behavioural, and social risk factors for CVD. Recently, there has been emerging interest in psychological risk factors for CVD with a particular focus on psychosocial stress. There has been accumulating evidence that psychosocial stress plays a role in the pathogenesis of CVD (Dimsdale, 2008; Hjemdahl, Rosengren, & Steptoe, 2011; Neylon et al., 2013; Steptoe & Kivimäki, 2013). Systematic reviews are in agreement that psychosocial stress predicts CVD incidence in initially healthy populations independent of standard risk factors (Everson-Rose & Lewis, 2005; Kuper, Marmot, & Hemingway, 2005). For the purposes of this literature review, psychosocial stress will be divided into three distinct categories: external stressors, negative emotional disorders, and acute stress triggers. Additionally, in this literature review I will also evaluate the evidence for the role of psychosocial stress in the prognosis of those already diagnosed with CVD.

22

1.4 Stress and cardiovascular disease: External stressors In this section I will seek to review the literature looking at associations between external stressors and CVD risk. Firstly, I will describe studies that have focused on broad composite measures of perceived life stress and chronic stress burden. Following this, I will describe associations between more specific types of psychosocial stress and CVD risk. These include caregiver stress, financial and work stress, and social isolation and loneliness. Where possible, results of systematic reviews and meta-analyses will be reported. 1.4.1 External stressors: Perceived stress and chronic stress burden The INTERHEART Study (Yusuf et al., 2004) examined the association between psychosocial stress over the previous 12 months and MI incidence in a standardised casecontrol study. Psychosocial stress was a composite self-report measure comprising stress at work and home, financial stress, the occurrence of major life events, lack of perceived control, and depression. This association was assessed in 15,152 MI cases and 14,820 CHD-free matched controls in 52 countries representing every inhabited continent. Results indicated that higher levels of psychosocial stress increased the risk of MI almost threefold after controlling for a range of traditional CVD risk factors, as well as geographic region. This association was seen in both men and women of all ages in all regions of the world. Andersen and colleagues (Andersen, Diderichsen, Kornerup, Prescott, & Rod, 2011) prospectively examined the association between major life events across the lifespan and incident CHD in 8,738 participants from the Copenhagen City Heart Study. There was no significant association between major life events and incident CHD. The authors put the lack of association down to the measurement of major life events, arguing that it may not 23

be a measure of chronic stress. They also argue that the 16-year follow-up period was too wide a timespan for the stress to have a meaningful effect on cardiac health. A meta-analysis carried out in 2012 examined the association of perceived stress and incident CHD (Richardson et al., 2012). Six (n=118,696) of the 23 potentially relevant articles met the criteria for section indicating that many of the studies examining this association were not of adequate quality. Meta-analysis revealed that high levels of perceived stress were associated with a moderately increased risk of incident CHD. However, the studies included in the meta-analysis differed in terms of covariates included in the models. All studies controlled for age, blood pressure, smoking, and cholesterol. Only three studies controlled for social factors such as SES, and only one study controlled for psychological factors such as depression and anxiety. A recent study prospectively examined the independent effects of individual-level stressors and neighbourhood-level stressors on incident CHD in a large sample from the Multi-Ethnic Study of Atherosclerosis with a 10.2 year follow-up period (Kershaw et al., 2015). Individual-level stressors included financial, work, relationship, and health-related stress. Neighbourhood-level stressors included neighbourhood safety and violence, social cohesion, and aesthetic quality. Higher individual-level stressors were linearly associated with incident CHD (n=6678). However, neighbourhood-level stressors were non-linearly associated with incident CHD, with medium levels of neighbourhood stress having a higher CHD risk (49%) than high levels of neighbourhood stress (27%). The authors find this result difficult to interpret and put it down to a stress measurement issue. Associations specifically between psychosocial stress and stroke incidence have also been described. Truelsen and colleagues (Truelsen, Nielsen, Boysen, & Grønbaek, 2003) prospectively examined associations between self-reported stress and stroke incidence

24

and fatality 13 years later in 12,574 men and women from the Copenhagen City Heart Study. Self-reported stress was measured in terms of stress intensity and frequency pertaining to feelings of tension, nervousness, impatience, anxiety, or sleeplessness. Results indicated that high stress frequency and intensity were associated with almost a doubled risk of fatal stroke compared to low stress after controlling for a number of traditional risk factors. But, self-reported stress was not associated with non-fatal stroke after adjustment for these covariates. The authors posit that the lack of significant association between stress and non-fatal stroke may be in part due to differences in CVD risk profiles amongst participants. A number of studies have also examined associations between psychosocial stress and both CHD and stroke incidence combined. Iso and colleagues looked at associations between perceived stress measured at baseline and stroke and CHD mortality in 73,424 initially disease-free Japanese men and women, with a follow-up of 580,378 person-years (Iso et al., 2002). Japanese women with high levels of perceived stress had a two-fold higher risk of death from stroke and CHD compared to those who reported low stress after adjusting for known cardiovascular risk factors. However, the same association was not observed in Japanese men. Stressful life events and social strain were measured at baseline in 82,000 women from the Women’s Health Initiative (Kershaw et al., 2014). After a follow-up period of 18 years, higher levels of stressful life events and social strain were associated with higher incident CHD and stroke. These associations were attenuated and became non-significant after adjustment for behavioural (e.g. smoking, dietary intake) and biological (e.g. hypertension, diabetes) CVD risk factors. The lack of association reported here lends support to the argument put forward by Andersen and colleagues (2011) that there was

25

too wide a timespan (18 years) between the stress exposure and the cardiovascular event for stress to have a meaningful effect on cardiovascular health. In the Hispanic Community Health Study, chronic stress burden, but not perceived or traumatic stress, was associated with a higher prevalence of CHD and stroke prevalence in 5313 men and women of mixed Hispanic/Latino ethnic backgrounds (Gallo et al., 2014). Additionally, chronic stress burden was associated with a higher prevalence of known CVD risk factors such as type 2 diabetes and hypertension in those free from CVD. Associations between stress and subclinical CVD have also been reported. Life stress (a composite measure of childhood trauma, negative life events, daily hassles, and job strain) was found to be associated with increased arterial stiffness, but not carotid atherosclerosis, in 650 participants from the Netherlands Study of Depression and Anxiety after controlling for many cardiovascular risk factors (Bomhof-Roordink et al., 2015). The studies outlined above focused on associations between broad composite measures of stress or perceived stress and CVD risk. Strengths of these studies include large sample sizes, with a number of studies being carried out across different cultures and ethnicities. On the whole these studies controlled for a large number of biological and behavioural cardiovascular risk factors. Overall, the evidence from these studies suggests that psychosocial stress is a significant risk factor for CVD incidence and mortality. However, a number of studies did not find such associations. How stress was conceptualised in these studies may be partially responsible for the lack of significant findings. The two studies that reported non-significant associations measured stress in terms of stressful or major life events (Andersen et al., 2011; Kershaw et al., 2014) whereas the other studies used either measures of perceived stress, or composite measures of a number of stress factors. What both these studies also have in common is the long follow-up length. Andersen and 26

colleagues (2011) argue that major life events may have a short-term effect on CVD risk, and if the life events occurred many years before the cardiovascular event this would explain why there is only a very weak association reported. 1.4.2 External stressors: Caregiver stress I will now describe research which has focused on specific types of external chronic stressors and their associations with cardiovascular risk. The chronic stress of caregiving for an elderly, ill, or disabled loved-one has been found to be associated with poor health and premature mortality (Schulz & Beach, 1999). Lee and colleagues examined 54,412 CVD-free women from the Nurse’s Health Study (Lee, Colditz, Berkman, & Kawachi, 2003). Information about caregiver stress was measured at baseline, and reports of incident CHD were collected throughout the four year follow-up period. Caregiving for an ill or disabled spouse for ≥ 9 hours per day was associated with an increased risk of incident CHD after adjusting for numerous cardiovascular risk factors. Interestingly, caregiving for an ill parent or other relative was not associated with higher CHD risk. This indicates that the high level of care required when taking care of a spouse may be more of a stressor and therefore increase CHD risk in women. Another study examining the effects of spousal caregiving strain on CVD risk found that high strain was associated with a 23% higher covariate-adjusted Framingham stroke risk score in both male and female caregivers (n=716) (Haley, Roth, Howard, & Safford, 2010). However, there was no association between caregiving strain and Framingham CHD risk scores (n=607). Capistrant and colleagues (Capistrant, Moon, Berkman, & Glymour, 2012) examined the association between spousal caregiving stress and CVD risk in 8,472 CVD-free participants from the Health and Retirement Study. Long-term spousal caregiving, defined as ≥14 hours of care per week measured in two consecutive

27

biennial questionnaires, was associated with a two-fold risk (hazard ratio=1.95) of CVD onset, but only in white individuals. Caregiving stress has also been associated with known CVD risk factors. Roepke and colleagues (Roepke et al., 2012) found that the duration of care in caregivers of those with Alzheimer’s disease was associated with increased carotid intima-media thickness independent of risk factors. This indicates that caregiving stress may increase CVD risk through atherosclerotic burden. Dementia caregivers have also been found to have higher levels of plasma IL-6 and D-dimer compared to sex-matched non-caregiving controls (von Känel et al., 2006). In terms of caregiving stress, the evidence does suggest that this type of chronic stress increases overall CVD risk, as well as levels of known CVD risk factors. Research indicates that the spousal caregiving is linked with increased CVD risk suggesting that caring for a spouse is a larger stressor than caring for another relative with a disability or illness. Studies in this area have largely focused on spousal caregiving. Further research should focus on CVD risk in other types of caregiver stress, such as CVD risk in carers/parents of sick children, or caregiver burden in mental illness. Interestingly, research in this area also indicates that ethnicity is an important factor in the association between caregiving stress and CVD risk, and therefore should be adjusted for in studies of this kind. Duration of care also appears to be important, indicating a cumulative effect of this type of chronic stress on CVD risk. 1.4.3 External stressors: Work stress, financial stress, and social isolation Work stress is the most widely studied external stressor. The work stress literature has been largely dominated by the ‘demand-control’ or ‘job strain’ model in which a combination of highly demanding work and low control conditions elicits stress in the 28

workplace (Karasek & Theorell, 1990). A systematic review examining work-related psychosocial factors and development of CHD found that there was moderate evidence that high demand, and a combination of high job strain and low social support at work (iso-strain), were associated with increased CHD risk in men (Eller et al., 2009). This finding was in men only as studies involving women were too few at the time to draw any meaningful conclusion. Pejtersen and colleagues (Pejtersen, Burr, Hannerz, Fishta, & Hurwitz Eller, 2015) updated this systematic review and meta-analysis with the results of 11 new studies examining work-related psychosocial factors and the development of CHD. The main result of this meta-analysis was that the ‘control’ element of job strain explained excess risk for MI amongst the selected studies (44 studies in total). However, results also revealed that a large amount of the selected studies (42/44) lacked sufficient power to detect a meaningful excess risk of MI. The authors also posit that the overwhelming focus on psychosocial stress models such as the job strain model make it difficult to paint a clear picture of what psychosocial factors at work are affecting CVD risk. A recent overview of systematic reviews carried out in this field confirmed the overwhelming focus on the job strain model in psychosocial stress research. Based on the evidence to date, the authors of this overview concluded that there is modest to moderate evidence for an association between psychosocial work stress and CVD risk in men (Fishta & Backé, 2015). The most compelling evidence for an association between work stress and CVD risk comes from a recent systematic review of the evidence from 27 studies from Europe, Asia, and the United States (n= >600,000) (Kivimäki & Kawachi, 2015). Results from this review found that work stress, with a focus on job strain and long working hours, was associated with a 10-40% excess risk of incident CHD and stroke, independent of conventional risk factors such as age, sex, and SES. They also reported associations 29

between work stress and type 2 diabetes, but not with cancer or chronic pulmonary obstructive disorder, which suggests outcome specificity in terms of work stress effects on health. A recent meta-analysis of the same magnitude (25 studies, n= >600,000) found that long working hours (≥55 hours per week) compared to standard working hours (3540 hours per week) were associated with an increased risk of CHD and stroke incidence (Kivimäki et al., 2015), after controlling for age, sex, and SES. The association between longer working hours and stroke was stronger than the association between working hours and CHD and demonstrated a dose-response association. Although distinct from work stress, financial stress has also been associated with CVD risk. A Swedish study reported that men without a cash margin (i.e. the ability to raise approximately £1000 in one week if an unexpected situation were to occur) had an increased risk of incident CVD after adjusting for relevant covariates (Carlsson et al., 2014). This link between financial strain and incident CVD was not present for women. Thus, we see that there is evidence suggesting that both work and financial stress are associated with increased CVD risk. However, these associations have been largely reported in men. Further research is needed in female samples to elucidate the effects work stress in this population. Also, overuse of the job strain model in studies assessing associations between work stress and CVD risk may be hampering our ability to assess what other elements of work stress are important. Future research should include other work-related variables, such as long working hours. Social isolation is another external stressor that has been associated with CVD progression. A meta-analysis of nine prospective cohort studies in CHD-free populations revealed that social isolation and loneliness were associated with a 50% excess risk of CHD on average (Steptoe & Kivimäki, 2012). A more recent meta-analysis of 11

30

longitudinal CHD and eight longitudinal stroke studies found that poor social relationships, defined as social isolation or loneliness, were associated with a 29% increase in the risk of incident CHD and a 32% increase in the risk of stroke (Valtorta, Kanaan, Gilbody, Ronzi, & Hanratty, 2016). 1.4.4 External stressors: Summary In sum, many external stressors have been associated with increased CVD risk, incident CVD, and CVD mortality. However, there are a number of issues to consider when interpreting the evidence. There are differences in the way stress is conceptualised across studies which potentially affect the associations reported with CVD. For example, measuring stressful life events rather than measuring broad composite measures of stress, or focusing on the job strain model rather than taking a wider approach to work stress, seems to attenuate the association between stress and CVD. Additionally, the timing between measurement of stress and measurement of cardiovascular health appears to be of importance to results. Studies with longer durations between these measurements have reported null findings (Andersen et al., 2011; Kershaw et al., 2014). In support of this, Nielsen and colleagues (Nielsen et al., 2006) reported that significant associations between high levels of perceived stress and CHD were attenuated as follow-up continued. This implies that psychosocial stress may have a relatively short-term effect on CVD incidence. Another prevalent issue in studies measuring associations between psychosocial stress and CVD is choice of covariates included in analyses. Most studies tend to adjust for well-established cardiovascular risk factors such as age sex, smoking, cholesterol, hypertension, etc. However, some fail to adjust for known social and behavioural cardiovascular risk factors. Therefore, the extent to which the stress-CVD link is associated with different behavioural, clinical, and social risk factors is difficult to interpret. Despite the problems listed above, most studies examining the associations 31

between external stressors and CVD tend to report at least modest to moderate associations after controlling for traditional risk factors, providing support for the role of these types of stressors in the development of CVD. 1.5 Stress and cardiovascular disease: Negative emotional disorders In this section I will review the literature examining associations between negative emotional disorders and CVD risk. Firstly, I will describe studies that have focused on psychological distress as measured by the 12-item General Health Questionnaire (GHQ) (Goldberg, 1992). The GHQ provides quite a comprehensive measure of psychological distress consisting of items that capture depressive symptoms, anxiety, social dysfunction, and loss of confidence. Following this, I will describe studies that have focused specifically on associations between depressive symptoms and CVD risk. Studies examining associations between anxiety and CVD risk will then be outlined. 1.5.1 Negative emotional disorders: Psychological distress Hamer and colleagues (Hamer, Molloy, & Stamatakis, 2008) examined data from 6,576 healthy men and women from the Scottish Health Study and revealed associations between baseline psychological distress and CVD events 7.2 years later. However, this association was only significant when adjusting for age and sex and did not survive the addition of behavioural cardiovascular risk factors including smoking, alcohol intake, and physical activity. A meta-analysis of 10 large prospective cohort studies from the Health Survey for England revealed an association between psychological distress and CVD mortality in 68,222 people who were initially disease-free (Russ et al., 2012). This association remained after adjustment for a number of relevant covariates including SES, body mass index (BMI), smoking status, alcohol intake, physical activity, blood pressure, and diabetes status. Another study using data from the Health Survey for England 32

examined associations between psychological distress and CVD mortality in 66,500 initially disease-free men and women and found that a one category increase in GHQ scores predicted increased stroke mortality (hazard ratio=1.18) and CHD mortality (hazard ratio=1.24) at a median follow-up time of 7.9 years (Lazzarino, Hamer, Stamatakis, & Steptoe, 2013). These associations were found to be strongest in the lowest SES categories. Similar covariates were adjusted for as in the meta-analysis carried out by Russ and colleagues (2012), with the absence of certain behavioural factors such as physical activity and alcohol intake. 1.5.2 Negative emotional disorders: Depression and anxiety There are a number of well-conducted systematic reviews and meta-analyses that show that depression is an independent risk factor for CVD (Dhar & Barton, 2016), and evidence suggests that as depressive symptoms worsen risk of developing CHD increases (Glassman & Shapiro, 1998). Nicholson and colleagues (Nicholson, Kuper, & Hemingway, 2006) carried out a meta-analysis of 21 aetiological studies examining associations between depression and future CVD. Together, these 21 studies comprised 124,509 participants and 416 cardiac events. Over a mean follow-up period of 10.8 years, results revealed an 80% higher risk of developing or dying from CHD in those with depression at baseline. Adjusting for other cardiovascular risk factors resulted in marginal reductions in relative risk. A later meta-analysis examining 28 studies confirmed the findings of Nicholson and colleagues (Van der Kooy et al., 2007). Sixteen of these studies examined CVD-free populations at baseline and found an increased risk of CVD in those who reported depressive symptoms at baseline (risk estimate = 1.57). The authors reported that clinically diagnosed major depression showed the greatest risk for the development of

33

CVD, equalling the risk of smoking and diabetes. A recent meta-analysis of 30 prospective cohort studies (n=893,850) revealed a pooled relative risk of 1.30 for both CHD and MI incidence in those with depression (Gan et al., 2014). Interestingly, the pooled relative risk for both CHD and MI was stronger with a follow-up period of less than 15 years (relative risk = 1.36) compared with follow-up periods of 15 years or longer (relative risk = 1.09). The most recent studies examining the role of depression in the aetiology of heart disease are in agreement with results from previous meta-analyses. In a study of 3572 men and women who had experienced an acute MI, 48% of the women and 25% of the men reported a lifetime history of depression (Smolderen et al., 2015). In a recent study using data from the Netherlands Study of Depression and Anxiety, there was a significant association between depression and new-onset CVD over a six year period in 2,510 initially CVD-free participants (Seldenrijk et al., 2015). Cox regression models revealed that having depression more than doubled the risk of developing new onset CVD (hazard ratio = 2.30). Anxiety has also been found to be an independent risk factor for CHD. Roest and colleagues carried out a meta-analysis of 20 studies reporting on anxiety and incident CHD over a mean follow-up period of 11.2 years in 249,846 individuals (Roest, Martens, de Jonge, & Denollet, 2010). The authors found people high in anxiety were at an increased risk of CHD (hazard ratio = 1.26) and cardiac death (hazard ratio = 1.48) independent of biological, social, and behavioural cardiovascular risk factors. There was no association between anxiety and nonfatal MI. A recent large meta-analysis of 37 studies (n=1,565,699) examining associations between anxiety and new onset CVD found that anxiety was associated with a 52% increase in risk of CVD (Batelaan, Seldenrijk, Bot, van Balkom, & Penninx, 2016). Anxiety was also associated with an increased risk 34

of MI of 38%, and a 74% increased risk for stroke. Adjustment for publication bias reduced the strength of all the reported associations. Although they remained significant the authors do not provide the attenuated hazard ratios. Therefore, results of this metaanalysis should be interpreted with that in mind. The latest meta-analysis examining anxiety and CVD risk included 46 cohort studies (n=2,017,276) and found that anxiety was associated with an increased risk of CVD mortality (41%), CHD (41%), stroke (71%), and heart failure (35%) (Emdin et al., 2016). However, in concurrence with Roest and colleagues (2010), anxiety was not associated with MI. The most recent study examining the role of anxiety in new onset CVD has reported that anxiety is a unique risk factor for stroke and MI in older primary care patients initially CVD-free (Stewart, Hawkins, Khambaty, Perkins, & Callahan, 2016). The authors examined the predictive value of anxiety and depression screening in 2,041 older primary care patients with a follow-up of eight years. Cox proportional hazards models revealed that a positive anxiety screen at baseline, but not a positive depression screen, was associated with a 54% increased risk of a CVD event in the first three years of follow up, after controlling for demographic and biological cardiovascular risk factors. However, after three years of follow-up this association disappeared. Conversely, a recent study found that depression, and comorbid depression and anxiety, was associated with new onset CVD, but anxiety alone was not (Seldenrijk et al., 2015). This indicates that inclusion of depressive symptoms as a covariate in research examining associations between anxiety and CVD is important. 1.5.3 Negative emotional disorders: Summary In sum, the evidence does seem to suggest that negative emotional disorders are associated with increased CVD risk, incidence, and mortality. However, as with research

35

on external stressors, some studies failed to adjust for known behavioural, social, and clinical risk factors. Failure to include some of these well-established risk factors could inflate the magnitude of associations reported between negative emotional disorders and CVD and this should be taken in to account when interpreting results. Interestingly, as seen with some external stressor types, the duration of follow-up in studies examining associations between depression and anxiety and CVD risk seems to be of importance to results. The relative risk of CVD in those with depression was 27% higher with a followup period of less than 15 years (Gan et al., 2014), and association between anxiety and CVD risk disappeared after three years of follow-up (Stewart et al., 2016). These findings support the notion that psychosocial stress may have a relatively short-term effect on CVD risk. 1.6 Stress and cardiovascular disease: Acute stress triggers Episodes of acute emotional stress have been shown to trigger adverse cardiovascular events in individuals with underlying CVD. In this section I will describe the literature examining associations between these acute stress triggers and cardiac events. Evidence for emotional triggering of cardiac events comes from both population-based studies and patient studies. Population-based studies have revealed that major events such as natural disaster, war, terrorist attacks, and major sporting events can trigger cardiac events in those with underlying CHD (Steptoe & Brydon, 2009). Natural disasters such as large-scale earthquakes, tsunamis, and hurricanes are recognised as acute stressors. A number of studies have described associations between natural disasters and increased rates of cardiac events and cardiac mortality. In the week following the Northridge Earthquake in California in 1994, hospital admissions for acute MI in the surrounding areas increased

36

by 35% (Leor & Kloner, 1996). This same earthquake was also found to increase rates of sudden cardiac death from the normal daily average of 4.6 (2.1) to 24 on the day of the earthquake (Leor, Poole, & Kloner, 1996). The Hanshin-Awaji earthquake in 1995 resulted in a 3.5 fold increase in hospital admissions for acute MI in the four weeks following the earthquake (Suzuki et al., 1997). Hospital admissions for ACS also increased in the three week period following the Great Eastern Japan Earthquake and tsunami in 2011 (Nozaki et al., 2013). Additionally, the incidence of sudden cardiac and unexpected deaths doubled in the four week period following this earthquake (Niiyama et al., 2014). The authors also reported significant associations between the rates of sudden death and seismic activity following the earthquake, indicating that the acute stress associated with fear of a repeated earthquake may have been a causal factor in these sudden deaths. Taken together, the studies provide evidence that acute stress brought about by an earthquake can trigger cardiac events. However, following the Loma Prieta earthquake in San Francisco in 1989, there was no increase in hospital admission for ACS (Brown, 1999). Steptoe and Brydon (2009) suggest that the timing of the earthquakes may provide an explanation for this disparity in findings. The Hanshin-Awaji and Northridge earthquakes struck in the early morning in winter, whereas the Loma Prieta earthquake occurred on an afternoon in the autumn. Susceptibility to acute MI is known to be raised in winter months, and in the early mornings (Elliott, 2001). The Great Eastern Japan Earthquake occurred on a spring afternoon, which is not in line with this argument. The subsequent occurrence of a large-scale tsunami and nuclear emergency may be the reason for this disparity. Large-scale natural disasters have also been shown to affect long-term cardiac health. In the three years following the Hanshin-Awaji earthquake, mortality from acute MI 37

increased by 14% (Nakagawa et al., 2009). In the six years following Hurricane Katrina, there was a more than three-fold increase in admissions for acute MI (Peters et al., 2014). Together, these studies suggest that as well as being acute stress triggers, natural disasters can result in chronic stress that affects the cardiovascular risk profile. This chronic stress may be to do with fear of recurrence of the disaster, bereavement, financial loss, forced migration, and general social upheaval. Research into acute stress triggers for cardiac events have also focused on the effects of war and terrorist attacks. During the Gulf War in 1991 incidences of acute MI and sudden cardiac death increased in response to Iraqi missile attacks in an area of Israel that was not hit by missiles, but was within hearing range of the explosions (Meisel et al., 1991). A number of studies have examined the cardiovascular effects of the terrorist attacks on the World Trade Centre in New York in 2001 (9/11). In the 60 days following these attacks, hospital admissions for acute MI increased by 49% in 16 New Jersey hospitals (Allegra, Mostashari, Rothman, Milano, & Cochrane, 2005), and increases in MI admissions were also observed in a Brooklyn hospital (Feng, Lenihanx, Johnson, Karri, & Reddy, 2006). However, a study of eight New York City hospitals found no acute increases in hospitalisation for cardiac events in the week following 9/11 (Chi, Speakman, Poole, Kandefer, & Kloner, 2003). Examining mortality data also found that there was no significant increase in cardiac deaths in New York in the months following the 9/11 attacks (Chi, Poole, Kandefer, & Kloner, 2003). These results are rather mixed. Holman and colleagues found that people who made subjective reports of high acute stress responses to the 9/11 attacks had a 53% increased incidence of cardiovascular events over the following three years, indicating that the degree to which the person found 9/11 stressful may account for the varying results across studies (Holman et al., 2008).

38

Although nowhere near as severe or traumatic as natural disasters or terrorist attacks, sporting events have also been found to have acute effects on cardiac health. During the 1998 World Cup, hospital admissions for acute MI increased by 25% in England on the day the English team lost to Argentina in a penalty shoot-out (Carroll, Ebrahim, Tilling, Macleod, & Smith, 2002). This effect extended to two days after the football match. A retrospective study examining the effects of football matches in England from 1994 to 1998 found that acute MI and stroke mortality was significantly increased (relative risk = 1.28) in men when the local football team lost at home (Kirkup & Merrick, 2003). Similar results have been reported in Germany with the incidence of acute cardiovascular emergencies increasing 2.66 times on World Cup match days involving the German team (Wilbert-Lampen et al., 2008). More specifically, research has shown that cardiovascular risk is increased in football fans only when the team in question loses. When Los Angeles played in the Superbowl and lost, deaths from CHD and acute MIs increased significantly (Kloner, McDonald, Leeka, & Poole, 2009). However, when Los Angeles played in the Superbowl and won all-cause mortality rates were reduced (ibid). Patient studies in acute stress trigger research have revealed that acute periods of intense, anger, stress, depression, and sadness can trigger coronary events (Steptoe & Brydon, 2009). A meta-analysis of five case-crossover studies (the gold standard of research in this area) revealed that the pooled relative risk of an ACS being preceded by a period of anger, sadness, or stress was 2.48 (Steptoe & Kivimäki, 2013). A recent meta-analysis examined nine independent case-crossover studies looking at associations between periods of intense anger and adverse cardiac outcomes. The authors concluded that there was an elevated risk of ACS, ischaemic and haemorrhagic stroke, and arrhythmia in the two hour period following an outburst of intense anger (Mostofsky, Penner, & Mittleman, 2014). However, these findings were more pronounced in people with higher underlying 39

cardiovascular risk who experienced frequent outbursts of anger in general. The most recent study examining the effects of anger on cardiac events is in keeping with the results of Mostofsky and colleague’s (2014) meta-analysis. Buckley and colleagues (Buckley et al., 2015) report results of a case-crossover study that revealed an increased relative risk (8.6) of experiencing an MI within 2 hours of experiencing very intense anger. Acute grief has also been shown to elevate risk of cardiac events. In a UK-based matched cohort study, the rate of MI and stroke in older adults who had recently lost their partners was increased almost two-fold, but only in the 30 days following the bereavement (Carey et al., 2014). In sum, the evidence supports the idea that intense emotional stress brought about by large-scale events or personal emotional experience can increase rates of cardiac events, in particular acute MI. But, as Steptoe and Brydon (2009) point out in their review, it is difficult to rule out alternative explanations for cardiac events following large-scale natural disasters, acts of terrorism, and sporting events. It is quite possible that disruption of health services at the time, or perhaps physical trauma or exertion, or even drinking too heavily at a football match, could have brought about the cardiac events in question. Although the evidence from patient studies indicates that intense emotions can trigger cardiac events, it is also possible that other factors are involved. 1.7 Stress and cardiovascular disease: Prognosis in those already affected As well as playing a role in the aetiology of CVD and the triggering of acute cardiac events, psychosocial stress can also worsen prognosis in those who already have CVD. In this section I will first describe literature examining the role of external stressors in CVD prognosis, with a particular focus on perceived stress, work stress, and the role of

40

social support. I will then discuss the role of negative emotional disorders in disease progression. 1.7.1 The role of external stressors in CVD prognosis In a large study of 4,204 acute MI patients, levels of perceived stress over the month preceding the MI were measured during hospitalisation. Patients with moderate to high perceived stress had increased 2-year all-cause mortality (hazard ratio = 1.42) compared with patients low in perceived stress after adjusting for conventional risk factors (Arnold, Smolderen, Buchanan, Li, & Spertus, 2012). Furthermore, patients with high/moderate perceived stress levels also had worse angina-specific quality of life one year after their initial MI. Similarly, high perceived stress scores measured during hospitalisation in 3,572 acute MI patients were associated with worse angina-related quality of life one month after the MI (Xu et al., 2015). The role of perceived stress in heart failure prognosis has also been examined but the authors reported that it was not significantly associated with event free survival in 81 heart failure patients (Alhurani et al., 2014). A recent systematic review and meta-analysis identified five papers derived from four different prospective cohort studies that examined associations between work stress and recurrent events in patients following their first cardiac event (Li, Zhang, Loerbroks, Angerer, & Siegrist, 2014). Meta-analysis (n=2,578) revealed that work stress increased the risk of future cardiac events by 65%. One of the studies included in the meta-analysis failed to find a significant association between work stress and further cardiac events in 292 female ACS patients in Sweden (Orth-Gomér et al., 2000). Interestingly, in these women marital stress was associated with a 2.9 fold increase in future cardiac events, even after adjusting for a large number of known cardiovascular risk factors. This may be because about a third of the sample was not in employment when baseline data were

41

collected. The authors also posit that women generally perceive spousal relationships as less supportive than men which may explain why marital stress, rather than work stress, was a predictor of future cardiac events in this study. Financial strain has also been associated with poor CHD prognosis. Financial strain over the previous year was measured in women who had been hospitalised for an ACS and was found to be associated with an almost threefold (hazard ratio = 2.76) risk of recurrent cardiac events after controlling for numerous potential confounders (Georgiades, Janszky, Blom, László, & Ahnve, 2009). Social support appears to have a protective role when it comes to CVD prognosis. Barth and colleagues carried out a systematic review and identified 20 prognostic papers examining associations between social support and CVD mortality suitable for inclusion in a meta-analysis (Barth, Schneider, & von Känel, 2010). Results indicated that patients with low functional support had an increased risk of both cardiac and non-cardiac mortality after adjustment for relevant risk factors (pooled hazard ratio = 1.59). High social support and strong social relationships have also been associated with better cardiovascular prognosis (Holt-Lunstad, Smith, & Layton, 2010). Thus, we see that as well as playing a role in CVD risk, external stressors play a role in CVD prognosis. However, this body of prognostic research is beset by similar issues seen in the CVD risk literature. Firstly, the way stress is conceptualised may be problematic – particularly in the prognostic literature relating to work stress. All the studies included in Li et al.’s (2014) meta-analysis conceptualised stress using the job strain model which means other psychosocial factors pertaining to work stress, and specifically the stress of returning to work after a cardiac event, were not considered. Secondly, the effects of some external stressors on CVD prognosis have not been examined. For example, the role of caregiver stress in CVD sufferers is yet to be explored. Associations between both marital 42

stress and financial strain and poor cardiovascular prognosis have been reported only in women, meaning that the relevance of these types of external stressors in men is as of yet unknown. Thirdly, studies measuring associations between external psychosocial stressors and CVD prognosis differ in terms of covariates adjusted for. A number of studies failed to control for important clinical and biological variables. Therefore, the extent to which stress affects prognosis may have been inflated in these studies. 1.7.2 The role of negative emotional disorders in CVD prognosis Negative emotional disorders have been associated with worse prognosis in CVD patients. Depression is prevalent and persistent in CHD patients and a comprehensive review has shown that 19.8% of acute MI survivors meet the criteria for major depression, while approximately 30% have mild-to-moderate depressive symptoms (Thombs et al., 2006). A number of meta-analyses have provided evidence for the link between depressive symptoms and worse prognosis in CVD patients. Van Melle and colleagues included 22 papers examining associations between depressives symptoms in acute MI patients and long-term cardiovascular prognosis in a meta-analysis (n=6,367) (Van Melle et al., 2004). The results indicated that MI patients with depression had more than a 2.5fold increase in cardiac mortality, and an almost two-fold risk for new cardiovascular events. Interestingly, neither follow-up duration nor method of measuring depression significantly affected the association between depression and mortality. A meta-analysis of 29 papers published in the same year also reported a two-fold increase of mortality in depressed patients in the two years after initial assessment (Barth, Schumacher, & Herrmann-Lingen, 2004). This association weakened after two years, but remained significant long-term.

43

In a 2006 meta-analysis of 34 prognostic studies, the pooled relative risk of all-cause or CHD mortality associated with depression was 1.80 (Nicholson et al., 2006). Interestingly, left ventricular function was only adjusted for in a small number of studies and inclusion of this covariate attenuated the relative risk by 48%. Although depression plays a role in CVD prognosis, this led the authors to suggest that depression was not yet an established independent risk factor for poor CHD prognosis as many studies failed to adjust for relevant risk factors. Meijer and colleagues identified 29 studies for inclusion in a meta-analysis examining the relationship between depression following the occurrence of an MI and cardiac prognosis (n=16,889) (Meijer et al., 2011). Similar to both meta-analyses carried out in 2004, the authors reported a 2.7-fold increased risk of cardiac mortality and a 1.6-fold increased risk of cardiac events in patients with post-MI depression. However, the strength of the association between depression and cardiac events decreased as follow-up duration increased – a finding also reported in Barth et al.’s (2004) meta-analysis. A recent meta-analysis sought to ascertain whether the cognitive/affective or somatic/affective symptoms of depression were more relevant for cardiovascular prognosis (de Miranda Azevedo, Roest, Hoen, & de Jonge, 2014). Thirteen prospective studies of 11,128 participants were included in the meta-analysis. In the fully adjusted analysis, somatic/affective depression symptoms, but not cognitive/affective symptoms, were associated with poor prognosis in CVD patients (hazard ratio = 1.19). There is evidence that anxiety is also associated with poorer prognosis in CVD patients. A 2010 meta-analysis of 12 studies comprising 5,750 MI patients reported associations between anxiety and cardiac mortality as well as new cardiac events independent of clinical variables, including depression (Roest, Martens, Denollet, & de Jonge, 2010). Roest and colleagues followed up this meta-analysis with a study examining associations 44

between generalised anxiety disorder (GAD) and adverse cardiac outcomes in MI patients with a 7-10 year follow-up period (Roest, Zuidersma, & de Jonge, 2012). Results from simple age and sex adjusted models showed that GAD was associated with an almost twofold risk of adverse events. Adjustment for various other clinical factors, including depression, did not affect the magnitude of the association greatly. However, the authors did not adjust for any social or behavioural factors. A systematic review of studies examining the role of worry and GAD in cardiovascular health found that three studies had reported associations between GAD and poorer prognosis in CHD patients, even after adjusting for depression (Tully, Cosh, & Baune, 2013). However, a year later Tully and colleagues carried out a meta-analysis on five studies examining the role of GAD in CHD patients and reported no significant associations (Tully, Cosh, & Baumeister, 2014). The latest meta-analysis in the area of anxiety and CVD prognosis included 44 articles examining prospective associations between anxiety and mortality in CHD patients (n=30,527) (Celano et al., 2015). After adjusting for a number of covariates, anxiety was not associated with mortality or poorer outcomes in CHD patients. The authors performed sensitivity analyses and found that when they separated the samples into post-ACS patients and stable CHD patients, the risk of poorer outcomes in anxious stable CHD patients was significantly elevated after adjusting for a number of relevant covariates. There were no significant increases in outcome risk in anxious post-ACS patients. In summary, the evidence suggests that negative emotional disorders play a role in prognosis in those already with CVD. Three meta-analyses to date have reported 2 to 2.5fold increases in risk of future cardiac events and mortality in CHD patients with depression (Barth et al., 2004; Meijer et al., 2011; Van Melle et al., 2004). However, the largest meta-analysis carried out so far (Nicholson et al., 2006) found that many studies 45

failed to adjust for relevant risk factors such as smoking and BMI, leading to inflated associations between depression and prognosis in CVD patients. More than 50% of patients suffering from depression or anxiety will also suffer from a comorbid depressive or anxiety disorder (Hirschfeld, 2001). Therefore, failure to adjust for symptoms of anxiety in many of these studies could also lead to inflated risk estimates. Adjusting for symptoms of depression seems to be more commonplace in prognostic studies measuring anxiety in CHD patients. This may be why the results of meta-analyses in this field are a little more mixed. Another reason for the mixed results seen in the prognostic metaanalyses related to anxiety may be failure to define samples correctly, i.e. separate stable CHD patients from post-ACS patients who are likely more symptomatic (Celano et al., 2015). Nevertheless, the literature suggests that both depression and anxiety play a significant role in CVD prognosis, but more work is needed with both well-adjusted statistical models and well-defined patient samples. 1.8 Chapter summary Overall the evidence suggests that psychosocial stress contributes significantly to the aetiology of CVD, CVD mortality, and CVD prognosis in those already affected. External life stressors, depression and anxiety, and intense periods of acute stress all seem to play a role in cardiovascular health. However, all studies in this area of research have been either cross-sectional or longitudinal prospective observational studies, meaning that these studies provide evidence for associations between stress and CVD, but are not able to establish causality. Results in this research area have been mixed and this is probably due to a number of methodological factors. On the whole, studies in this field tend to be well-powered and well-designed. But, there are issues with how stress is conceptualised and measured that

46

may affect the results of these studies. As mentioned earlier in this literature review, measuring life events rather than how people perceive life events as stressful, or focusing on the ‘job strain’ model rather than taking a wider approach to measuring work stress has likely affected results in the external stressor literature. Additionally, failure to adjust for covariates relevant to the development of CVD may have resulted in inflated associations between psychosocial stress and CVD outcomes. In general, most studies tend to adjust for traditional risk factors. But, health behaviours, social, and psychological factors known to be relevant to cardiovascular risk are often not controlled for. One interesting issue that emerges from the stress-CVD literature is duration of time between measurement of stress and cardiovascular event. Longer follow-up durations seem to weaken associations between psychosocial stress and cardiovascular risk and this is seen in external stressor research (Andersen et al., 2011; Kershaw et al., 2014), and both depression (Barth et al., 2004; Gan et al., 2014; Meijer et al., 2011) and anxiety (Stewart et al., 2016) research. What this suggests is that psychosocial stress likely has cumulative effects that lead to biological alterations that increase CVD risk over time, and that stress needs to be sustained in order to have a long-term effect. The lack of significant findings in the studies with long follow-up durations implies that perhaps the stress had dissipated (i.e. major life events), or the depression or anxiety symptoms had been dealt with or had waned. Nevertheless, this body of research does support the role of psychosocial stress in cardiovascular disease. The next step is to increase our understanding of the underlying biological mechanisms and pathways that link psychosocial stress with CVD. Then we may be able to devise targeted interventions to prevent psychosocial stress from developing into disease. In the next chapter I will discuss the role of a specific biological

47

pathway, the hypothalamic-pituitary-adrenal axis, in the link between psychosocial stress and CVD.

48

Chapter 2 Literature review: The role of the hypothalamic-pituitary-adrenal axis and the corticosteroid receptors 2.1 Introduction In this chapter I will define and describe the hypothalamic-pituitary-adrenal (HPA) axis and its role in the stress response. I will then provide evidence for associations between chronic stress and dysregulation of the HPA axis, and associations between dysregulation of the HPA axis and CVD risk and prognosis. By doing this, I hope to show how dysregulation of the HPA axis might be one of the biological pathways linking psychosocial stress and CVD. I will then introduce the corticosteroid receptors and define and describe their role in the stress response. I will argue that stress-related modulation of these receptors, resulting in reduced glucocorticoid sensitivity, might be one mechanism through which HPA axis dysregulation is brought about. Thus, the aim of this chapter is to highlight the role of stress-related HPA axis dysregulation in CVD, and to provide evidence for the role of the corticosteroid receptor in HPA axis dysregulation. 2.2 Potential pathways linking psychosocial stress and CVD There are a number of pathways through which psychosocial stress may contribute to the pathophysiology of CVD. One possibility is that the relationship between stress and CVD may be mediated through behavioural pathways. Psychosocial stress can influence CVD risk indirectly by increasing more adverse health behaviours (Steptoe & Kivimäki, 2012). A prospective cohort study (n=7,066) examining stress-related changes in health behaviours found that individuals with high levels of perceived stress were less likely to quit smoking over time, more likely to be sedentary, and less likely to keep alcohol consumption within the recommended limits (Rod et al., 2010). Psychosocial stress and 49

particularly depression has been associated with poorer adherence to medication in CHD patients (Gehi, Haas, Pipkin, & Whooley, 2005). Additionally, psychological distress has been associated with poor cardiac rehabilitation attendance in CHD patients (Glazer, Emery, Frid, & Banyasz, 2002). There is substantial evidence from both observational and laboratory studies suggesting that there are direct pathophysiological links between psychosocial stress and CVD. Psychosocial stress factors have been associated with increases in autonomic and endothelial dysfunction, increased systemic inflammation, upregulated cellular adhesion, and also promotion of a pro-thrombotic state (von Känel, 2012). Of particular relevance to this PhD is the association between psychosocial stress and alterations in HPA axis activity. The HPA axis is the major neuroendocrine system in humans that is activated during times of stress and incorporates a major part of the stress response. Before describing associations between psychosocial stress, alterations in HPA axis function, and the development of CVD, I will provide a brief overview of the stress system and the stress response. 2.3 The stress system and the stress response When homeostasis is threatened, or perceived to be so, the stress response is initiated. The stress response is an adaptive response that brings about changes in the sympathoadrenal-medullary (SAM) system and the HPA axis which then go onto induce cardiovascular, metabolic, and immune changes that serve to protect the body from stress (Brotman, Golden, & Wittstein, 2007). The neural circuitry that initiates the stress response is mainly located in the hypothalamus and the brain stem. This circuitry includes corticotropin releasing hormone (CRH) neurons of the paraventricular nucleus of the

50

hypothalamus, and the locus coeruleus-norepinephrine (LC-NE) system in the pons and medulla (Tsigos & Chrousos, 2002). The SAM system is comprised of the sympathetic nervous system (SNS) and the adrenal medulla. During times of stress, the SAM system is activated by the LC-NE system. Firstly, the LC-NE system releases epinephrine into the brain which results in heightened alertness and a decrease in functions such as sleeping and eating (Brotman et al., 2007). This system also stimulates the hypothalamus which activates the SNS and results in the secretion of epinephrine and norepinephrine from the adrenal medulla. The release of these catecholamines results in increased heart rate, blood pressure, blood viscosity, and inflammation (Brotman et al., 2007). That is, in times of acute stress the SAM system is responsible for initiating the ‘fight or flight’ response, readying the body for any injury that may occur. During times of stress (see Figure 2.1), the HPA axis is activated by CRH from the paraventricular nucleus of the hypothalamus, which then leads to the release of adrenocorticotropic hormone (ACTH) from the anterior pituitary gland. ACTH then stimulates the release of glucocorticoids from the adrenal cortices, as well as some mineralocorticoids and androgens. Cortisol is the neuroendocrine end-point of the HPA axis and is the main circulating glucocorticoid in humans. Cortisol is a pleiotropic hormone. It has central energy-conserving effects as well as regulatory effects on the metabolism of protein, glucose, and fat for energy release. Cortisol exerts an immunomodulatory effect inhibiting the stress-related release of a number of inflammatory cytokines such as interleukin (IL) -6, IL-1, and tumour necrosis factor-α (TNF-α) (Kaltsas, Zannas, & Chrousos, 2012). Cortisol also increases blood pressure in times of stress via vasoconstriction (Girod & Brotman, 2004). In addition, cortisol exerts

51

a regulatory effect on itself suppressing the release of its biological precursors (CRH and ACTH) thus forming a negative feedback loop.

Figure 2.1. The hypothalamic-pituitary-adrenal axis

During times of stress, there is a lot of cross-talk between the SAM system and the HPA axis. For example, while catecholamines released by the adrenal medulla serve to stimulate secretion of IL-6 (März et al., 1998), the HPA axis serves to inhibit the release of pro-inflammatory cytokines (Kaltsas et al., 2012). Moreover, IL-6 has also been found to stimulate activation of the HPA axis independent of CRH release (Bethin, Vogt, & Muglia, 2000). In sum, both systems serve to regulate each other, and a number of inflammatory mediators. However, when these systems become dysregulated, there can be adverse cardiovascular consequences. Although both systems are interrelated, this PhD will predominantly focus on the causes of HPA axis dysregulation and its implications for cardiovascular health.

52

2.4 Stress-related HPA axis activity HPA axis activity increases in response to stress, resulting in increased levels of cortisol in humans, in order to exert the metabolic, cardiovascular, and anti-inflammatory effects required to maintain what Sterling and Eyer referred to as ‘allostasis’ (Sterling & Eyer, 1988). Allostasis is the process of achieving stability through a number of short-term adaptive physiological changes (McEwen & Wingfield, 2003). Alterations in HPA axis activity in response to stress is just one example of allostasis. Other examples include alterations in catecholamine levels, cytokine levels, heart rate, and blood pressure (McEwen & Wingfield, 2010). When allostasis is called upon too often, or is not managed efficiently, the demand of all these adaptive processes on the body can take its toll. This is referred to as ‘allostatic load’ or ‘allostatic overload’ (McEwen, 2000). Allostatic load can be interpreted as the ‘wear and tear’ of certain biological systems when faced with either too much stress, or failure to adapt to stress biologically, i.e. not ‘turning off’ the stress response when it is no longer required (McEwen, 2007). As the stress response is sustained over time, the biological ‘wear and tear’ on the body and brain can lead to the development of pathology and illness (McEwen, 2008). Figure 2.2 illustrates four different conditions that lead to allostatic load (McEwen, 1998). For the purposes of this PhD the ‘physiological response’ referred to in the figure represents the cortisol stress response. Box (1) illustrates the normal cortisol response to stress, where cortisol increases, the increase is sustained for an appropriate amount of time in order for cortisol to exert its effects, and then the response is turned off. Box (2) represents repeated stress ‘hits’ from different stressors meaning that the cortisol stress response is frequently being triggered. Box (3) represents repeated stress hits from the same type of stressor and failure to habituate to that stressor. Box (4) represents a prolonged cortisol response to stress due to a delayed or failed shut down of the stress 53

response. Finally, Box (5) represents an inadequate or blunted cortisol response to stress. The prolonged (Box 4) and inadequate (Box 5) cortisol responses are likely what occurs after many repeated stress ‘hits’ (Box 2), or failure to adapt to stress (Box 3). A prolonged increase in cortisol after stress indicates that the hormone is unable to exert its selfregulatory function. An inadequate cortisol response to stress means that cortisol cannot exert its regulatory effects on inflammation, metabolism, and the cardiovascular system, potentially leading to hyperactivity of other stress-related mechanisms (e.g. increased inflammation).

Figure 2.2. Four types of allostatic load. Box (1) represents the normal cortisol stress response. Box (2) represents cortisol responses to repeated different stress hits. Box (3) represents lack of adaptation of cortisol to a similar stressor. Box (4) represents a prolonged cortisol response due to a delayed shutdown. Box (5) represents an inadequate cortisol stress response. 54 Adapted from McEwen (1998).

2.5 Diurnal HPA axis activity Under basal, unstressed conditions, the HPA axis shows marked diurnal patterning and levels vary substantially throughout the day (Figure 2.3). These patterns can be observed with repeated plasma samples, but are more commonly assessed noninvasively with assays of free cortisol in saliva. The circadian rhythm of the HPA axis is regulated by the suprachiasmatic nucleus in the hypothalamus, which responds to levels of light in the environment, and then goes on to stimulate the release of CRH from the neurons in the paraventricular nucleus (Spiga, Walker, Terry, & Lightman, 2014). Cortisol is at high levels on waking, followed by a rise that reaches a peak approximately 30 minutes after waking. This is referred to as the cortisol awakening response (CAR). There is then a subsequent decline across the day (the cortisol slope), with cortisol reaching its nadir at around midnight (Adam & Kumari, 2009).

Figure 2.3. The diurnal nature of HPA axis function. Under basal (unstressed) conditions, cortisol secretion is characterised by a circadian rhythm. The decline of cortisol across the day is referred to as the cortisol ‘slope’.

55

As well as dysregulation of the HPA axis manifesting as prolonged or inadequate cortisol responses to stress, dysregulation also occurs at the basal, unstressed level. Dysregulation of basal HPA axis function can cause alterations in the diurnal cortisol rhythm that take the form of blunted or heightened CARs, as well as flatter or steeper cortisol slopes (Adam & Kumari, 2009). These flatter cortisol slopes can be driven by lower waking cortisol levels, higher evening cortisol levels, or both. Another proxy for HPA axis dysregulation is area under the curve (AUC) which is an estimate of average cortisol exposure across the day (Adam & Kumari, 2009). 2.6 Psychosocial stress and dysregulation of diurnal HPA axis function A growing body of evidence suggests an association between psychosocial stress and dysregulation of diurnal HPA axis function. A number of different types of psychosocial stress have been shown to be associated with alterations in diurnal HPA axis indices to date including acute psychosocial stress, a number of types of chronic stress, as well as stress-related disorders such as depression (Collomp et al., 2016). 2.6.1 Psychosocial stress and the CAR The CAR can be determined by either calculating simple change scores between cortisol levels at waking and levels at 20-45 minutes after waking, or by calculating the AUC, or the overall volume, of cortisol released over this waking period (Clow, Thorn, Evans, & Hucklebridge, 2004). Chida and Steptoe carried out a meta-analysis of 62 studies that examined associations between different psychosocial stress factors and the CAR (Chida & Steptoe, 2009). These psychosocial stress factors included job stress, general life stress (including perceived stress), depression, anxiety, posttraumatic stress disorder (PTSD), fatigue or burnout, as well as positive psychological factors such as positive affect and optimism. The authors separated studies into those that examined the CAR calculated 56

using simple change scores (CARi) and those that examined the CAR by calculating the AUC over the waking period (CARAUC). Meta-analysis revealed that the CARi was positively associated with levels of job stress and general life stress. Interestingly there were negative associations between the CARi and fatigue or burnout. Similarly, the CARAUC was found to be associated with higher general life stress. When the authors limited the meta-analysis to include only studies with high methodological quality scores, the positive association between the CARAUC and general life stress remained, but a negative association between the CARAUC and PTSD also emerged. These results indicate that different psychosocial stress factors have differing effects on the CAR. More recent research has replicated the negative associations between the CAR and people with burnout. Oosterholt and colleagues reported a blunted CAR in clinical and non-clinical burnout patient groups compared to healthy controls (Oosterholt, Maes, Van der Linden, Verbraak, & Kompier, 2015). Negative associations between the CAR and PTSD have also been replicated with a blunted CAR being reported in 24 adolescent girls with diagnoses of PTSD (Keeshin, Strawn, Out, Granger, & Putnam, 2014). Although Chida and Steptoe’s meta-analysis reported associations between enhanced CAR and general life and job stress, some research carried out since the meta-analysis reports contradictory associations. Academic stress was found to be associated with a reduced CAR in 42 healthy young men compared to 21 age-matched men not undergoing examinations (Duan et al., 2013). These reductions in CAR were negatively correlated with perceived stress and anxiety levels. Cropley and colleagues reported blunted CAR in school teachers with high levels of work-related rumination compared to teachers with low levels (Cropley, Rydstedt, Devereux, & Middleton, 2015). High levels of perceived stress have also been associated with blunted CAR in 64 healthy men and women 57

(Gartland, O’Connor, Lawton, & Bristow, 2014). Conversely, enhanced CAR has been reported in 24 healthy adults who had experienced early life trauma (Lu et al., 2013). This finding was later replicated in 58 people who had experienced childhood trauma (Lu, Gao, Huang, Li, & Xu, 2016). Collectively, what these results indicate is that different types of stress potentially have differing effects on the CAR. Other factors may come into play such participant age, sex, or clinical status. Additionally, the timing and duration of the type of stress is probably important. Whether or not participants were going through a particularly stressful period at the time of data collection may have affected results. Acute anticipatory stress has been shown to result in an increase in cortisol levels after awakening (Wetherell, Lovell, & Smith, 2015). Working mothers reporting high job strain and high parenting stress have been found to have enhanced CAR increases on workdays compared with non-workdays (Hibel, Mercado, & Trumbell, 2012). What these results suggest is that stress-related situational factors affect the CAR. Therefore, future studies should seek to measure and adjust for these factors. Interestingly, depression has also been found to have a varying association with the CAR. There has been a large amount of research carried out on CAR profiles in depression. A recent systematic review of the literature concluded that depression is associated with both a heightened and a blunted CAR (Dedovic & Ngiam, 2015). The authors suggest that this discrepancy might be related to depression severity. In fact, one study examining basal HPA axis function in depression described an inverted U-shaped association between depression and CARAUC (Veen et al., 2011). This non-linear association has been replicated in a much larger sample from the Netherlands Study of Depression and Anxiety (Wardenaar et al., 2011). It is possible that this non-linear association was the reason why there was no association between depression and the CAR in Chida and Steptoe’s (2009) 58

meta-analysis. However, a recent case-controlled study examining associations between depression and morning cortisol in older adults did not observe this U-shaped association between morning HPA axis function and depression (Rhebergen et al., 2015). 2.6.2 Psychosocial stress and the cortisol slope The cortisol slope is a measure of cortisol decline across the day and can be calculated in a number of ways. Typically, where a number of cortisol samples have been provided across the day a line of best fit is applied to each individuals’ data points using linear regression and the slope of this line is used as an estimate of the cortisol slope across the day (Adam & Kumari, 2009). The CAR sample (waking +30 minutes) is generally not included in these calculations and the slope is based on the first sample of the day taken upon waking. Flatter cortisol slopes have been associated with a number of different stressors and negative emotional disorders. In a 2007 meta-analysis, Miller and colleagues examined the effects of chronic stress on a number of diurnal cortisol parameters (Miller, Chen, & Zhou, 2007). They identified 119 papers (n=8,521) studying a number of different types of chronic stress including combat/war experience, abuse/assault, bereavement, caregiving stress, natural disasters, and job loss. Meta-analysis revealed that exposure to chronic stress was significantly associated with a flatter diurnal rhythm, as well as significantly lower morning cortisol levels and higher afternoon/evening levels which were likely resulting in the flattened slope. Research carried out in this area since this meta-analysis has largely corroborated this result. In a large study (n=1,694) of men and women from the National Study of Daily Experiences, a greater frequency of daily stressors was associated with a flatter diurnal cortisol slope (Stawski, Cichy, Piazza, & Almeida, 2013). This association remained 59

robust after adjustment for negative affect, and clinical factors also. Perceived stress in the home has also been associated with a flatter cortisol slope in men, but not in women (Sjörs, Ljung, & Jonsdottir, 2014). Work stress has been associated with flatter cortisol slopes. In a large occupational cohort (n=2,126), effort-reward imbalance was found to be related to flatter cortisol rhythm throughout the day (Liao, Brunner, & Kumari, 2013). However, this association was modest after adjustment for a number of demographic factors. Family-to-work spillover, i.e. the extent to which work infringes on your family life, has also been associated with flatter cortisol slopes (Zilioli, Imami, & Slatcher, 2016). In a study of 98 older adults from the Brain Health Substudy of the Baltimore Experience Corps Trial, those deemed as socioeconomically disadvantaged had flatter cortisol slopes across the day (Agbedia et al., 2011). Early life adversity has also been associated with flattened diurnal cortisol rhythms in children and adolescents (Koss, Mliner, Donzella, & Gunnar, 2016; McLachlan et al., 2016). However, a previous study reported no difference in cortisol slope between women who had experienced early life adversity and matched controls who had not (Gonzalez, Jenkins, Steiner, & Fleming, 2009). Negative emotional disorders, in particular depression, have also been associated with aberrant cortisol rhythm throughout the day. In a sample of 257 Swedish men and women, depression was found to be associated with flatter diurnal cortisol slopes (Sjögren, Leanderson, & Kristenson, 2006). Flatter cortisol slopes have also been reported in women with major depressive disorder (Jarcho, Slavich, Tylova-Stein, Wolkowitz, & Burke, 2013), as well as in adolescents who have had recent episodes of major depressive disorder (Doane et al., 2013). Negative emotions such as sadness, loneliness, and high reports of general distress were also associated with flattened rhythms in this adolescent sample (ibid). More severe depressive symptoms have also been associated with more 60

pronounced flattening of the cortisol slope (Hsiao et al., 2010). In disagreement with previous research, a recent study reported no difference in cortisol slope between depressed and non-depressed individuals (Booij et al., 2015). It is possible that this study may have been underpowered to detect significant differences between groups (n=15 per group). Conversely, associations between positive psychosocial factors and steeper cortisol declines across the day have been reported, lending support for the associations between negative stress factors and flatter cortisol slopes. Social support and positive coping styles were associated with steeper cortisol rhythms across the day (Sjögren et al., 2006). There have also been associations reported between high levels of positive affect, such as feelings of alertness and activeness, and steeper diurnal cortisol slopes (Hoyt, Craske, Mineka, & Adam, 2015). 2.6.3 Psychosocial stress and cortisol AUC The cortisol AUC is not a measure of the circadian variation of cortisol but instead reflects the average levels of cortisol secreted throughout the day (Adam & Kumari, 2009). Nevertheless, associations between cortisol AUC and psychosocial stress factors have also been reported. As with research looking at the CAR, results from AUC research have been varied. Miller and colleagues (2007) meta-analysis cited earlier also elicited significant associations between exposure to chronic stress and a higher daily volume of cortisol output. However, more recent research has produced mixed results. Examination stress has been linked with reduced cortisol AUC in healthy young men (Duan et al., 2013). Job strain has also been linked with altered cortisol AUC. In 104 healthy adults, higher job strain was associated with higher cortisol AUC (Maina, Bovenzi, Palmas, & Filon, 2009). Conversely, a more recent study reported lower total cortisol AUC in older

61

adults with higher levels of job strain from the Multi-Ethnic Study of Atherosclerosis Stress Study (Rudolph et al., 2016). The authors posit that the age of the participants may be the reason for the discrepancy between these two studies. Interestingly in a sample of 68 healthy younger adults there was no association between job strain and cortisol AUC (Maina, Palmas, & Filon, 2007). However the cortisol AUC was significantly higher on working days compared with non-working days. This indicates that, like the CAR, situational factors have influence on the AUC. Temporality may also be an issue affecting results in these studies. In their meta-analysis, Miller and colleagues (2007) showed that as time since the stress exposure increased the strength of the association between stress and cortisol AUC decreased. Alterations in cortisol AUC have also been reported in depression. In 45 female caregivers of stroke survivors, higher levels of depressive symptoms were associated with lower cortisol levels across the day (Saban, Mathews, Bryant, O’Brien, & Janusek, 2012). In a study of 401 men and women in Canada, lower cortisol concentrations across the day were associated with symptoms of depression, psychological distress, and burnout (Marchand, Durand, Juster, & Lupien, 2014). Conversely, elevated cortisol AUC has been reported in 57 depressed individuals compared to healthy controls (Dienes, Hazel, & Hammen, 2013). In a meta-analysis of 20 studies examining salivary cortisol in depression, the results suggested that salivary cortisol levels are generally increased in patients with a depressive disorder (Knorr, Vinberg, Kessing, & Wetterslev, 2010). However, using meta-regression the authors found that the difference in salivary cortisol levels observed was probably associated with age and intra-assay variability of the cortisol kits, rather than depression scores. Low social support has been associated with higher cortisol AUC in healthy students (Heaney, Phillips, & Carroll, 2010), whereas in older adults low social support was associated with a reduced cortisol AUC (Piazza, 62

Charles, Stawski, & Almeida, 2013).In this same older sample, higher levels of negative affect were positively associated with cortisol AUC. What these results suggest is that, like CAR, alterations in cortisol AUC related to depression are mixed and are possibly related to other factors like age, temporality, and other psychosocial factors such as social support and affect. 2.7 Psychosocial stress and dysregulation of stress-related HPA axis activity Acute psychosocial stress induced in the laboratory leads to activation of the HPA axis and a subsequent increase in cortisol levels (Dickerson & Kemeny, 2004). There is a body of evidence suggesting that exposure to chronic stressors can bring about alterations in the magnitude of cortisol responses to acute stress. 2.7.1 Exposure to chronic stress and early life adversity The most comprehensive meta-analysis to date examining the effects of chronic psychosocial factors on cortisol responses to acute stress in the laboratory revealed that positive psychological traits, i.e. openness, spirituality, self-esteem, and positive coping style, were associated with reduced cortisol stress reactivity in the laboratory (Chida & Hamer, 2008). However, there were no significant associations between negative stressrelated factors and laboratory-induced cortisol stress responses due to inconsistency between studies. This lack of association may have been down to the nature of the chronic stressor or the duration between stress exposure and acute stress testing (Miller et al., 2007). Perhaps the positive psychological factors were associated with cortisol responses in the laboratory because they are stable traits rather than transient stress factors that can dissipate over time. Much of the research looking at the effects of chronic stress exposure on acute cortisol stress reactivity has focused on early life adversity. In healthy adults with no 63

psychopathology, those with a history of moderate to severe childhood maltreatment exhibited blunted cortisol responses to psychosocial stress in the laboratory compared to those with no experience of maltreatment (Carpenter et al., 2007). In a follow-up study the authors replicated these findings in a non-clinical sample of women and found that those who had experienced childhood physical abuse had blunted cortisol responses to laboratory stress compared to those who had not (Carpenter, Shattuck, Tyrka, Geracioti, & Price, 2010). Similarly, in a study of 80 healthy men and women exposed to a psychosocial laboratory stress, those who had high exposure to adverse childhood events (n=33) had significantly blunted cortisol responses to the stress tasks compared to those with no exposure to adverse events (Elzinga et al., 2008). Pre-stress cortisol values did not differ between groups. In a highly cited study, Heim and colleagues compared four different groups of women (n=49) on cortisol and ACTH responses to acute psychosocial stress in the laboratory (Heim, Ehlert, & Hellhammer, 2000). One group had current major depression and had experienced childhood abuse, one group was free from depression but had experienced childhood abuse, one group had current major depression and had not experienced childhood abuse, and the control group had experienced neither depression nor abuse in childhood. The results indicated that after the acute stress protocol, abused women with current major depression exhibited significantly higher cortisol responses to stress compared with the other three groups. In terms of ACTH responses, both groups of abused women, regardless of depression status exhibited significant increases compared to non-abused depressed women and controls. These findings are in contrast with those of Suzuki and colleagues who found that cortisol responses to stress were blunted in those who had experienced childhood trauma, regardless of depressive status (Suzuki, Poon, Papadopoulos, Kumari, & Cleare, 2014). 64

Goldman-Mellor and colleagues compared three different groups of healthy men and women from the Whitehall II cohort (n=543) (Goldman-Mellor, Hamer, & Steptoe, 2012). Two of the groups had experienced early life stress whereas one had not (control group). Of the early life stress groups, one had a history of recurrent psychological distress over the previous 20 years, whereas the other group did not. Following an acute stress laboratory protocol, those who had experienced both early life stress and recurrent psychological distress had blunted cortisol responses to stress compared with the control group. Conversely, similar to the findings of Heim and colleagues, those who had experienced early life stress with little or no history of ongoing distress had elevated baseline cortisol levels and prolonged cortisol responses to stress compared to the control group. These results differ from the earlier studies mentioned above. This may be because in these earlier studies the ‘healthy’ participants who had experienced early childhood adversity may have had underlying depressive symptomatology or psychological distress that was not taken into account. The discrepancies between results in this area could also have to do with the way early childhood adversity is defined. Stress involving threat to the physical self or trauma are known to elicit different HPA axis responses compared to stress that threatens the social self (Miller et al., 2007). 2.7.2 The effects of depression Cortisol stress reactivity has been found to be dysregulated in depression. In a small metaanalysis of seven studies (n=196), those with major depressive disorder were found to have prolonged cortisol responses compared to non-depressed individuals indicating delayed shutdown of the stress response (Burke, Davis, Otte, & Mohr, 2005). However, within this meta-analysis, older patients and more severely depressed patients were found to have blunted cortisol reactivity to acute stress, particularly when the laboratory session was in the afternoon. The results of a more recent study partially mirror those of this meta65

analysis. Amongst 351 adolescents from the TRAILS cohort, Booij and colleagues found that adolescents with recent-onset major depressive problems had prolonged responses to a laboratory-based stress protocol (Booij, Bouma, de Jonge, Ormel, & Oldehinkel, 2013). However, those who had persistent or recurrent depression throughout adolescence had blunted cortisol responses to the same stress protocol. These results suggest that initially, depressive symptoms might enhance (prolong) the cortisol stress response, but over time responsivity diminishes possibly due to repeated stress hits. This may be why blunted cortisol responses to stress were seen in older and more severely depressed patients in Burke et al.’s (2005) meta-analysis. Amongst a sample of older people (n=68, >55y) with elevated cardiovascular risk, those who were clinically depressed were found to have blunted cortisol responses to acute laboratory stress compared to their non-depressed counterparts (Taylor et al., 2006). Similarly, a recent study showed that in a large older sample (n=725, 50-65y) from the Dutch Famine Birth Cohort Study, higher symptoms of depression and anxiety were associated with blunted cortisol stress reactivity in the laboratory (de Rooij, 2013). This finding is in support of the notion that older age, and therefore perhaps longer exposure to depression and anxiety throughout the lifespan, results in diminished cortisol reactivity to stress. A recent study examining cortisol stress reactivity in youth depression (n=115, 9–16y) found that depressive symptoms were associated with higher cortisol responses to a socially evaluated cold-pressor test, but only in boys (Lopez-Duran et al., 2015). This lends further support that age, and exposure to depression, plays a role in the association between depression and cortisol stress reactivity. Overall, the evidence suggests that psychosocial stress factors and negative emotional disorders are associated with dysregulation of both basal and stress-related HPA axis function. Different stress types seem to exert different effects on the direction of 66

dysregulation of the HPA axis. For example, both heightened and blunted CARs have been reported across different types of stress. This also applies to cortisol AUC as well as cortisol stress reactivity. Factors that appear to influence the direction of dysregulation are age, temporal issues, and also the severity of the stressor. However, in terms of cortisol slope, flatter cortisol slopes seem to be uniformly associated with stress-related factors. As mentioned previously, cortisol is a pleiotropic hormone that exerts regulatory effects on energy release, cardiovascular function, and the release of a number of proinflammatory cytokines, as well as regulating its own release via a negative feedback loop. Stress-related dysregulation of the HPA axis may then have further reaching biological implications that could promote the development of a number of diseases, including CVD. In the next section I will provide evidence for the role of HPA axis dysregulation in CVD. 2.8 HPA axis dysregulation and CVD In a comprehensive review, Girod and Brotman lay out the ways in which the HPA axis is important for cardiovascular function and reduction of CVD risk (Girod & Brotman, 2004). Firstly, they note that a normally functioning HPA axis ‘primes’ the body for stress by preparing the metabolic, cardiovascular, haemostatic and autonomic components of the stress response required for the experience of everyday stress. Secondly, they outline the ‘suppressive’ role of cortisol in that it prevents inflammation and tissue repair processes from exceeding required levels and resulting in damage to the self. Thirdly, cortisol is known to play a role in insulin sensitivity, lipid production, and fat accumulation (Peckett, Wright, & Riddell, 2011). Based on these three roles of the HPA axis outlined above, dysregulation of the axis and abnormal cortisol secretion could therefore negatively alter cardiovascular risk (Girod & Brotman, 2004).

67

2.8.1 Diurnal HPA axis function in CVD Standard observational methods have revealed associations between dysregulation of basal diurnal HPA axis activity and progression of CVD. High levels of cortisol reactivity in the hour after waking have been found to be positively associated with intima media thickness of the artery carotis communis in women (Eller, Netterstrøm, & Allerup, 2005; Eller, Netterstrøm, & Hansen, 2001). Morning levels of cortisol have been found to be elevated in men who have moderate to severe coronary atherosclerosis (Troxler, Sprague, Albanese, Fuchs, & Thompson, 1977). In the CARDIA study, there was a significant cross-sectional association between a flatter cortisol slope across the day and higher levels of coronary artery calcification in 718 healthy middle-aged adults (Matthews, Schwartz, Cohen, & Seeman, 2006). In the Multi-Ethnic Study of Atherosclerosis Stress Study, a unit increase in coronary calcium was associated with a 1.77% flatter decline in cortisol in 464 older men and women (Hajat et al., 2013). In 1,866 healthy participants from the Rotterdam Study, higher cortisol AUC values were associated with an increased number of atherosclerotic plaques in the carotid arteries (Dekker et al., 2008). Dysregulated diurnal HPA axis function has been reported in clinical cohorts also. Patients with CHD have been found to have flatter diurnal cortisol slopes compared to healthy controls (Nijm, Kristenson, Olsson, & Jonasson, 2007). However, this finding was not replicated by Bhattacharyya and colleagues who examined cortisol slopes in patients with CAD compared to those without (Bhattacharyya, Molloy, & Steptoe, 2008). The CAR has also been found to be blunted in CVD patients (Vreeburg et al., 2009). Interestingly, dysregulation of the CAR seems to vary according to disease severity. CHD patients who had a history of MI had a more blunted CAR compared to CHD patients who had no previous MI (Merswolken, Deter, Siebenhuener, Orth-Gomér, & Weber, 2013). 68

Thus, the evidence suggests that dysregulation of basal HPA axis function is associated with markers of cardiovascular risk, as well as being characteristic of CHD itself. This implies that HPA axis dysregulation may be one of the biological pathways through which psychosocial stress causes the development of CVD. The impact of stress on the pathophysiology of CVD is also likely to be mediated in part by mild chronic systemic inflammation (Steptoe & Kivimäki, 2013). The role of inflammation in atherosclerosis is well established (Hansson & Hermansson, 2011) and markers of low grade inflammation have been associated with higher risk of CVD (Danesh et al., 2004). Seeing as glucocorticoids serve to regulate inflammation, it is likely that dysregulation of the HPA axis contributes to chronic systemic inflammation characteristic of CVD. In fact, in the cross-sectional study where Nijm and colleagues showed that flatter cortisol slopes were seen in CHD patients compared to healthy controls, they also reported that levels of evening cortisol (which were the driving force behind the flattened cortisol rhythm) were strongly correlated with serum levels of IL-6 and C-reactive protein (CRP) (Nijm et al., 2007). An important study has shown that dysregulation of the HPA axis not only plays a role in the development of CVD, but is also associated with cardiovascular mortality. Kumari and colleagues examined diurnal cortisol patterns in 4,047 civil servants from the Whitehall II cohort and assessed mortality data over a follow-up period of 6.1 years (Kumari, Shipley, Stafford, & Kivimaki, 2011). The results showed that flatter cortisol slopes were associated with increased risk of all-cause mortality, but that this association was mainly driven by an increased risk of cardiovascular death. These results indicate that dysregulation of diurnal cortisol secretion is related to CVD mortality in originally disease-free individuals. To date, no one has examined the role of diurnal HPA axis dysregulation in the prognosis of those who already have advanced CVD. 69

Therefore, the first study of this PhD presented in Chapter 3 will examine whether pre-surgical diurnal cortisol profiles can predict adverse clinical outcomes in patients with advanced heart disease. 2.8.2 Cortisol stress reactivity in CVD Evidence for the role of HPA axis dysfunction in CVD also comes from laboratory studies of cortisol stress reactivity. Dysregulated cortisol responses to stress have been associated with elevated CVD risk factors. As mentioned before, the role of systemic inflammation in atherosclerosis is well established. Inflammation increases in response to acute stress challenges (Steptoe, Hamer, & Chida, 2007). In a laboratory-based acute stress study healthy middle-aged participants were divided into cortisol responders and cortisol nonresponders. Following the stress protocol, cortisol non-responders had higher levels of plasma IL-6 and a greater IL-1 receptor antagonist (IL-1Ra) response to stress compared with cortisol responders (Kunz-Ebrecht, Mohamed-Ali, Feldman, Kirschbaum, & Steptoe, 2003). This suggests that an adequate cortisol response to stress is required to regulate the inflammatory stress response. Those with blunted cortisol stress reactivity (i.e. the non-responders) had both increased systemic inflammation (IL-6), an increased inflammatory stress response (IL-1Ra), as well as lower heart rate variability, which are all factors associated with the development of CVD (Kunz-Ebrecht et al., 2003). Interestingly, cortisol responders to acute laboratory stress have been found crosssectionally to have increased levels of significant coronary artery calcification (Agatston score ≥100) after adjustment for a number of traditional risk factors (Hamer, O’Donnell, Lahiri, & Steptoe, 2010). Since interpretation of causality in cross-sectional data can be problematic, the authors decided to carry out a prospective follow-up of this study. They examined coronary artery calcification progression over the three year follow-up period

70

and found an association between higher, or more prolonged, cortisol stress reactivity and rate of calcification progression (Hamer, Endrighi, Venuraju, Lahiri, & Steptoe, 2012). In older initially healthy men and women, cortisol stress reactivity in the laboratory was found to be associated with higher incident hypertension at three year follow-up after adjusting for a number of clinical factors (Hamer & Steptoe, 2012). The results of the four aforementioned studies provide conflicting results. On the one hand blunted cortisol stress reactivity is associated with a number of CVD risk factors, and on the other, heightened cortisol stress reactivity is associated with increased coronary artery calcification (a sub-clinical marker of atherosclerosis), and incident hypertension. It is possible that age is a factor in the discrepancy between these results. In the earlier study by Kunz-Ebrecht and colleagues, the sample was comprised of healthy middle-aged participants. In the studies carried out by Hamer and colleagues, the samples were comprised of healthy older adults. Age is known to be a strong regulatory factor of cortisol secretion (Veldhuis, Sharma, & Roelfsema, 2013). Nevertheless, the results of these studies provide evidence that dysregulation of the cortisol stress response, regardless of direction, is associated with adverse cardiovascular and atherosclerotic factors. Cortisol responses to acute laboratory stress have also been measured in CHD patients. Thirty patients who had recently experienced a first-time ACS underwent a psychosocial stress protocol comprising anger recall and arithmetic. Compared with age-matched healthy controls, the CHD patients had blunted cortisol responses to stress, even after adjusting for confounding factors such as smoking or medication use (Nijm et al., 2007). A very recent study has replicated these findings. In 91 participants who underwent the Trier Social Stress Test (TSST) in the laboratory, those who had CHD (n=46) had blunted cortisol stress reactivity compared to those who were CHD-free (Waller et al., 2016). This 71

group difference remained significant even after adjustment for cardiovascular medication use. These findings are in line with the results of a population based study which showed that Lithuanian men had significantly lower cortisol responses to acute psychosocial stress in the laboratory compared to men from Sweden (Kristenson et al., 1998). Men from Lithuania have been shown to have a four-fold risk for CHD mortality, more atherosclerotic plaques, increased intima-media thickness, and higher levels of carotid artery stiffness compared to men from Sweden (Kristenson et al., 2000). Taken together, these observational and laboratory-based studies suggest that dysregulation of the HPA axis, through changes in both the diurnal cortisol profile and cortisol stress reactivity, may increase CVD risk and progression. The evidence suggests that psychosocial stress factors and negative emotional disorders can bring about dysregulation of the HPA axis. Therefore, it is possible that dysregulation of the HPA axis may be one of the biological pathways through which psychosocial stress ‘gets under the skin’ and affects the pathophysiology of CVD. It is therefore important that we establish how psychosocial stress might bring about sustained changes in HPA axis function. One possible course is via changes in the sensitivity of the corticosteroid receptors. 2.9 The role of the corticosteroid receptors Cortisol exerts its effects by binding to its receptors – the glucocorticoid receptor (GR) and the mineralocorticoid receptor (MR). GRs are ubiquitously expressed around the body, whereas MRs are expressed only in selected tissues such as the kidney, colon, heart, and central nervous system (CNS). In their inactivated state, both receptors reside within the cell cytoplasm anchored in place by chaperone molecules. Once bound to cortisol, the receptor sheds its chaperone molecules and translocates into the cell nucleus (see Figure

72

2.4). Within the cell nucleus there are two distinct mechanisms of action through which the ‘activated’ receptor exerts effects on gene transcription. Firstly, the ligand-receptor complex can directly bind to glucocorticoid response elements (GREs) in target genes in order to enhance gene transcription. The activated receptor can also bind with negative GREs in order to inhibit gene transcription. Binding to GREs represents the classic model of corticosteroid receptor action (Bamberger, Schulte, & Chrousos, 1996) and allows cortisol to exert its regulatory effects. The second mechanism of action is largely of relevance to the anti-inflammatory effects of cortisol. A number of immune genes (e.g. IL-6, IL-2) do not have GREs yet their expression is suppressed by cortisol (Bamberger et al., 1996). This is because cortisol can also exert its effects by binding directly with transcription factors within the cell nucleus, such as Nuclear factor-κb (NF-κB) or activator protein-1 (AP-1), in order to down-regulate inflammatory gene transcription (Girod & Brotman, 2004). The MRs are referred to as Type I receptors. They have a high affinity for endogenous glucocorticoids (i.e. cortisol in humans, corticosterone in rats) and aldosterone (salt and water regulation) and are therefore thought to regulate basal activity of the HPA axis as well as the onset of the stress response (de Kloet, 1998). The GRs are referred to as Type II receptors. They have a high affinity for dexamethasone (a synthetic glucocorticoid) but a low affinity for endogenous glucocorticoids. Therefore, they are thought to be important in the regulation of the stress response when levels of endogenous glucocorticoids are high, and the subsequent shutdown of the cortisol stress response via the negative feedback loop of the HPA axis (Carvalho & Pariante, 2008). One explanation of HPA axis dysregulation may be diminished sensitivity of the corticosteroid receptors. With reduced receptor sensitivity, cortisol is no longer able to exert its regulatory effects successfully leading to a breakdown in the HPA axis negative feedback loop, and an 73

Figure 2.4. Translocation of the corticosteroid receptor into the cell nucleus. Once a glucocorticoid binds to its receptor, the receptor sheds its chaperone molecules (HSP90, p23) and translocates into the cell nucleus where it binds to GREs on target genes, or directly with transcription factors, in order to activate or repress gene transcription. HSP90: heat shock protein 90; p23: protein 23; GRE: glucocorticoid response element

increase in the intensity of the inflammatory response (Cohen et al., 2012). An increase in the intensity or duration of the inflammatory response has consequences for the development and progression of chronic inflammatory diseases such as CVD (Danesh et al., 2004). 2.9.1 What causes modulation of corticosteroid receptor sensitivity? There is a substantial amount of difference between individuals in corticosteroid receptor sensitivity (Quax et al., 2013) and there are a number of factors that modulate this sensitivity. Firstly, the extracellular and intracellular bioavailability of glucocorticoids will affect sensitivity of the corticosteroid receptors. For example, patients on long term exogenous treatment with synthetic glucocorticoids will quite often develop tissue74

specific glucocorticoid resistance (Oakley & Cidlowski, 2013). Circulating levels of corticosteroid-binding

globulin

(CBG)

may

influence

the

bioavailability

of

glucocorticoids (Bamberger et al., 1996). CBG is the major transporter protein for glucocorticoids and binds to approximately 80-90% of all circulating cortisol. The remaining 20% is comprised of albumin-bound and free cortisol. Only cortisol not bound to CBG is biologically active (Lewis, Bagley, Elder, Bachmann, & Torpy, 2005). Therefore, the amount of CBG in circulation will influence the amount of cortisol available to act on intracellular corticosteroid receptors. CBG levels are under complex regulatory control and exposure to inflammatory cytokines such as IL-6 and IL-1β have been shown to influence CBG secretion and messenger RNA (mRNA) levels (EmptozBonneton, Crave, LeJeune, Brébant, & Pugeat, 1997). CBG release is also known to be affected by psychosocial stress (Kumsta, Entringer, Hellhammer, & Wüst, 2007). Variations in the levels of 11β-hydroxysteroid dehydrogenase (11β-HSD) may also affect receptor sensitivity. 11β-HSD is an enzyme that can convert cortisol both to its active and inactive forms. 11β-HSD-1 converts cortisone, which is biologically inactive, to cortisol. 11β-HSD-2 oxidises cortisol into the inactive metabolite cortisone. Changes in the levels of these enzymes within the cell exert effects on the bioavailability of cortisol, thereby affecting corticosteroid receptor sensitivity (Oakley & Cidlowski, 2013). Increased levels of 11β-HSD-1 have been associated with an increase in GR sensitivity (Whorwood, Donovan, Wood, & Phillips, 2001). Interestingly, increased levels of 11β-HSD-2 have been found in the offspring of maternal Holocaust survivors who underwent severe trauma (Bierer et al., 2014). Corticosteroid receptor sensitivity may also be affected by the number of receptors in the cell, or the ‘hormone binding capacity’ of the cell (Bamberger et al., 1996). Lower cell receptor concentrations have been found in patients with depression (Pariante & Miller, 75

2001), and depression has been associated with decreased glucocorticoid senstivity (Pace, Hu, & Miller, 2007). Glucocorticoids themselves have been shown to bring about significant downregulation of corticosteroid receptors (Bamberger et al., 1996). This may be an adaptive function preventing tissue damage from overexposure to glucocorticoids, which may over time become maladaptive. The hormone binding affinity of the receptors also likely plays an important role in modulation of receptor sensitivity. Every receptor has a ligand-binding domain which is the area to which glucocorticoids bind. Coincubation with IL-2 and IL-4 has brought about alterations in the ligand-binding domains of human lymphocytes leading to reduced hormone binding affinity of the GR (Kam, Szefler, Surs, Sher, & Leung, 1993). What this indicates is that increased inflammation may bring about reduced corticosteroid receptor sensitivity through reducing hormone binding affinity. Differing ratios of splice variants of the corticosteroid receptors may also affect receptor sensitivity. The GR gene NR3C1 consists of nine exons which are subject to splicing, which gives rise to a number of splice variants of the gene, two of which are the GRα and GRβ isoforms (Quax et al., 2013). In isolation, GRα facilitates the action of glucocorticoids, whereas GRβ is inactive. However, when GRβ is co-expressed with GRα, GRβ inhibits the action of GRα which suggests that a higher GRβ:GRα ratio may lead to glucocorticoid resistance (Oakley & Cidlowski, 2013). The GRβ isoform is present in many cells, but is usually found in lower levels than the GRα isoform. However, cytokines have been found to influence the expression of GR splice variants. IL-2 and IL-4 were found to increase the expression of GRβ isoforms in peripheral blood mononuclear cells (PBMCs) by more than 100% (Leung et al., 1997). The proinflammatory cytokines TNF-α and IL-1β were shown to increase GRα expression by 150% while increasing GRβ by 350% in HeLA cells which express both isoforms 76

endogenously (Webster, Oakley, Jewell, & Cidlowski, 2001). The results of these studies indicate that inflammatory cytokines could bring about a decrease in corticosteroid receptor sensitivity through upregulation of the GRβ splice variant of the receptor. Individual variation in GR sensitivity may also be influenced by genetic difference. Functional polymorphisms of the GR gene have been shown to influence the effects of glucocorticoids. Individuals with the ER22/23EK polymorphism of the GR gene have been found to demonstrate glucocorticoid resistance, whereas individuals with the N363S single nucleotide polymorphism (SNP) have demonstrated enhanced GR sensitivity (Manenschijn, Van Den Akker, Lamberts, & Van Rossum, 2009). 2.9.2 How do we measure corticosteroid receptor sensitivity? Corticosteroid receptor sensitivity can be indirectly assessed both in vivo and in vitro. Assessment involves measuring associations between a specific input (e.g. different concentrations of synthetic glucocorticoids) and suppression of a specific output, such as ACTH or cortisol production, mitogen-induced lymphocyte proliferation, or lipopolysaccharide (LPS)–induced inflammatory cytokine production (Rohleder, Wolf, & Kirschbaum, 2003). These associations allow us to examine glucocorticoid sensitivity in peripheral blood cells thus providing us with a proxy measure of corticosteroid receptor sensitivity. Note that from now on the terms ‘glucocorticoid sensitivity’ and ‘corticosteroid receptor sensitivity’ will be used interchangeably. In vivo, the most widely used method to examine glucocorticoid sensitivity is the dexamethasone suppression test (DST) (Rohleder et al., 2003). This test involves peripheral administration (usually oral) of a low dose of the synthetic glucocorticoid dexamethasone which in theory should then suppress the release of ACTH from the pituitary via negative feedback. In turn, the release of cortisol should also be suppressed. 77

The results of the DST can be interpreted as an index of glucocorticoid sensitivity with non-suppression of cortisol release being indicative of diminished GR sensitivity (Ebrecht et al., 2000). In vitro assays have also been developed in order to examine glucocorticoid resistance within immune cells. In this assay the effect of dexamethasone on lymphocyte proliferation or production of inflammatory cytokines, both of which should be inhibited by glucocorticoids, is used as an index of GR sensitivity (Carvalho & Pariante, 2008; Rohleder et al., 2003). The most common assay used today was developed by DeRijk and colleagues who use LPS to stimulate the release of inflammatory cytokines in whole blood, or PBMCs isolated from whole blood, (DeRijk, Petrides, Deuster, Gold, & Sternberg, 1996). LPS is an endotoxin produced by gram-negative bacteria known to induce an inflammatory immune response from cells (Raetz & Whitfield, 2002). The inhibition of LPS-stimulated secretion of inflammatory cytokines by different concentrations of dexamethasone is used as an index of glucocorticoid sensitivity. Failure to inhibit, or partial inhibition, indicates reduced in vivo GR sensitivity and nonsuppression in the DST has been correlated with reduced dexamethasone-induced inhibition of lymphocyte proliferation in vitro (Carvalho & Pariante, 2008). Within this thesis, these in vitro assays will be referred to as glucocorticoid sensitivity assays. As dexamethasone has a high binding affinity for the GR, the DST and glucocorticoid sensitivity assays outlined above only provide a proxy measure of sensitivity of this specific receptor. The prednisolone suppression test (PST) has been developed which allows the evaluation of both the GR and the MR. Prednisolone is a synthetic glucocorticoid which is more similar than dexamethasone to cortisol and therefore binds to both the GR and the MR (Pariante et al., 2002). Thus, the inhibition of LPS-stimulated secretion of inflammatory cytokines by different concentrations of prednisolone provides 78

an indirect measure of GR and MR sensitivity. Prednisolone can also be used in vitro in glucocorticoid sensitivity assays. In vitro glucocorticoid sensitivity assays are usually performed using whole blood or using PBMCs isolated from whole blood. Whole blood allows for rapid measurement of peripheral glucocorticoid sensitivity in white blood cells. However, it does not account for differences in cell population ratios within the white blood cells which could influence variability within the sample being measured (Burnsides et al., 2012). Therefore, it is preferable to carry out these assays using specific isolated PBMCs such as lymphocytes or monocytes. One flaw of glucocorticoid sensitivity assays is that they are not tissue specific. These assays are carried out using whole blood or PBMCs meaning that the results give an indication of peripheral corticosteroid receptor sensitivity and cannot be extended to other tissues of interest, such as cardiac or brain tissue (Carvalho & Pariante, 2008). Also, measuring a small number of specific outcomes, such as LPS-induced IL-6 or TNF-α levels, means we are not examining all the wider effects of glucocorticoids (Quax et al., 2013). Therefore results of these assays should be interpreted with these issues in mind. There are other in vitro methods used to measure corticosteroid receptor sensitivity and receptor function. The number of receptors within cells and the hormone binding affinity of the receptors can be measured directly using a glucocorticoid binding assay (Chriguer et al., 2005). Corticosteroid receptor mRNA expression can be assessed in PBMCs and receptor protein levels can be measured directly using Western blot techniques and indirectly using cytosol binding (Carvalho & Pariante, 2008). This provides an indication of the number of receptors within the cells. Measuring the number of corticosteroid receptors is an indicator of glucocorticoid sensitivity, but does not provide information

79

about the biological effectiveness of the receptor (Quax et al., 2013). Examining the rate of translocation of the corticosteroid receptors into cell nuclei also provides another proxy of receptor function. 2.10 Psychosocial stress and the corticosteroid receptors To date, a number of studies have examined the effects of both chronic and acute stress on corticosteroid receptor sensitivity. Before describing this body of literature it is worth noting that the majority of research has focused on the sensitivity of the GR, with very little attention paid to the MR. I will first describe studies that have looked at associations between chronic stress and corticosteroid receptor sensitivity, and then move on to describe studies of the effects of acute stress on receptor sensitivity in the laboratory. 2.10.1 Chronic stress and corticosteroid receptor sensitivity Many types of chronic stressors, including negative emotional disorders, have been found to affect the sensitivity of GRs (See Table 2.1). Associations between job strain and glucocorticoid resistance have been reported. In a study measuring vital exhaustion in male industrial employees, those who were highly exhausted had reduced GR sensitivity compared to non-exhausted employees (Wirtz et al., 2003). Highly exhausted employees also had elevated levels of CRP. However, a recent study reported increased GR sensitivity and function in 12 men suffering from job-related exhaustion compared to 12 matched healthy controls (Menke et al., 2014). In 46 healthy school teachers, those who reported high levels of effort-reward imbalance at work had reduced GR sensitivity compared to those with low effort-reward imbalance (Bellingrath, Rohleder, & Kudielka, 2013). In a study assessing the effects of academic stress on glucocorticoid resistance in 11 healthy students, the authors compared glucocorticoid sensitivity in lymphocytes one hour before an examination and also on a control day during a holiday period (Sauer et 80

al., 1995). They found that academic stress resulted in a decrease in cortisol inhibition of lymphocyte IL-2 production, implying reduced lymphocyte sensitivity to cortisol. The authors posit that exposure of lymphocytes to increased cortisol levels during the preexam stress period may have resulted in a loss of GR sensitivity. This reduction in sensitivity could be an adaptive response to short-term hypercortisolism. However, the small sample size means that results should be interpreted with caution. Reduced lymphocyte sensitivity to cortisol has also been reported in elderly caregivers of dementia patients compared to elderly non-caregivers (Bauer et al., 2000). Miller and colleagues reported decreased dexamethasone suppression of LPS-induced IL-6 production in whole blood of parents of children with cancer compared to parents of healthy children, indicating reduced GR sensitivity to dexamethasone (Miller, Cohen, & Kim, 2002). These same parents also reported high levels of psychological distress, and had flatter cortisol slopes across the day. However, in a recent study there were no significant differences in GR protein levels or hydrocortisone suppression of LPSinduced IL-6 production from monocytes in adult caregivers of family members with glioblastoma compared with controls whose lives were free of major stressors (Miller et al., 2014). The authors posit that hydrocortisone could be acting on the MR which may be why there were no significant differences in the caregiver sample. Reduced GR sensitivity has also been reported in those suffering from emotional disorders. Women with major depressive disorder were shown to have diminished GR sensitivity compared to healthy controls, and this diminished sensitivity was associated with flatter diurnal cortisol slopes (Jarcho et al., 2013). A systematic review of 34 studies examining associations between early life stress, depression, and GR and MR sensitivity found that early life stress leads to reduced inhibitory feedback of the HPA axis via

81

Table 2.1. Studies examining the effects of chronic stress on corticosteroid receptor function Author/date

Sample

Sauer et al. (1995)

11 healthy students (6 female), mean age 19y

Bauer et al. (2000)

49 spousal caregivers of dementia patients (24 female), mean age 72y, 67 matched noncaregiver controls

Miller et al. (2002)

Wirtz et al. (2003)

Study design

Chronic stress type

GR/MR measurement protocol

Statistical test and covariates

Main findings

Differences in glucocorticoid sensitivity during examination period and holiday period Differences in glucocorticoid sensitivity between elderly caregivers and non-caregivers

Academic stress

Cortisol suppression of PHA-induced lymphocyte proliferation in isolated PBMCs

Spearman’s rank correlations; no covariates

Caregiver stress

DEX and cortisol suppression of PHAinduced lymphocyte proliferation in isolated PBMCs

ANOVA; no covariates

25 parents of children undergoing cancer treatment (mean age 36y), 25 matched controls with healthy children

Differences in glucocorticoid sensitivity between both groups of parents

Chronic psychological stress of having a child who is undergoing treatment for cancer

DEX suppression of LPSinduced IL-6, IL-1β, and TNF-α production in whole blood

ANOVA; baseline cytokine values

Parents of children with cancer had reduced GR sensitivity compared to parents of medically healthy children. They also had flatter cortisol slopes.

325 healthy adults (280 male), mean age 40y

Difference in glucocorticoid sensitivity between those who are nonexhausted, and highly exhausted

Vital exhaustion in industrial employees

DEX suppression of LPSinduced IL-6 production in whole blood

ANOVA; no covariates

Men who were highly exhausted had reduced GR sensitivity compared to those who were non-exhausted, but not those who were moderately exhausted.

Academic stress was associated with reduced glucocorticoid sensitivity, implying reduced sensitivity of the corticosteroid receptors. Caregivers had reduced glucocorticoid sensitivity compared to non-caregivers, implying reduced sensitivity of the corticosteroid receptors.

DEX = dexamethasone; DST = dexamethasone suppression test; ERI = effort-reward-imbalance; GLM= general linear model; GR = glucocorticoid receptor; IL-1β = interleukin-1β; IL-6 = interleukin-6; LPS = lipopolysaccharide; PHA = phytohaemagglutinin (stimulates lymphocyte proliferation); PBMC = peripheral blood mononuclear cell; TNF-α = tumour necrosis factor – α; TSST = Trier Social Stress Test; WC = waist circumference.

82

Table 2.1 (continued) Studies examining the effects of chronic stress on corticosteroid receptor function Author/date

GR/MR measurement protocol

Statistical test and covariates

ERI at work and the TSST

DEX suppression of LPSinduced IL-6 production in whole blood

ANCOVA; gender, BMI, depression scores

High levels of ERI were associated with decreased GR sensitivity both before and after acute stress, compared to low levels of ERI.

Difference in GR sensitivity and function between exhausted and non-exhausted men

Job-related exhaustion

GR sensitivity: DST in vivo, GR function: DEX induced gene expression

GLM, linear regression; age, BMI

Enhanced GR sensitivity and function in those suffering from exhaustion.

Difference in GR sensitivity between caregivers and noncaregivers

Caregiver stress

GR sensitivity: hydrocortisone suppression of LPSinduced IL-6 production in monocytes; monocyte expression of GR protein levels measured using flow cytometry

Generalised estimating equations; age, sex, ethnicity, education, smoking, alcohol intake, physical activity, WC.

Both groups had similar levels GR protein levels.

Sample

Study design

Bellingrath et al. (2013)*

46 healthy adults (29 female), mean age 50y.

Associations between chronic stress and GR responses to the TSST in the lab

Menke et al. (2014)

12 men suffering from job-related exhaustion (mean age 45y), and 12 matched healthy controls

Miller et al. (2014)

33 caregivers of relatives with cancer (21 female, mean age 54y) and 47 noncaregiving matched controls

Chronic stress type

Main findings

No difference in GR sensitivity between groups.

DEX = dexamethasone; DST = dexamethasone suppression test; ERI = effort-reward-imbalance; GLM= general linear model; GR = glucocorticoid receptor; IL-1β = interleukin-1β; IL-6 = interleukin-6; LPS = lipopolysaccharide; PHA = phytohaemagglutinin (stimulates lymphocyte proliferation); PBMC = peripheral blood mononuclear cell; TNF-α = tumour necrosis factor – α; TSST = Trier Social Stress Test; WC = waist circumference. *Although study includes an acute psychosocial stress measure, the relevant associations are between a measure of chronic stress and GR sensitivity

83

changes in GR and MR sensitivity, and the subsequent development of depression (von Werne Baes, de Carvalho Tofoli, Martins, & Juruena, 2012). However, within this review there was large methodological variation between studies which may have affected the conclusions drawn. Instead of measuring sensitivity of receptors, some studies have examined the number of corticosteroid receptors within each cell as a proxy for their sensitivity. Calfa and colleagues found that depressed patients had reduced GR numbers in PBMCs compared to healthy controls (Calfa et al., 2003). Together, this body of research indicates that chronic stress, including depression, results in decreased sensitivity of the GR. Pariante and colleagues posit that this diminished GR sensitivity brings about impaired feedback inhibition of the HPA axis, thus explaining the enhanced cortisol stress reactivity seen in major depression (Pariante, Thomas, Lovestone, Makoff, & Kerwin, 2004). As well as shutting down the cortisol stress response, the GR are also responsible for regulating the magnitude of the response. This means that diminished GR sensitivity could also explain the blunted cortisol stress reactivity observed in older and more severely depressed patients (Burke et al., 2005; Taylor et al., 2006). Interestingly, the MR appears to be slightly oversensitive in depressed patients (Young, Lopez, Murphy-Weinberg, Watson, & Akil, 2003). As the MR regulates basal activity of the HPA axis, altered MR sensitivity may have implications for the dysregulation of diurnal HPA axis activity brought about by depression. It is therefore likely that depression is characterised by an imbalance of both GR and MR sensitivity (de Kloet, DeRijk, & Meijer, 2007). This imbalance in sensitivity likely has consequences for levels of inflammation in the body also. In support of this, a number of the studies outlined above reported higher levels of inflammation and signs of HPA axis dysregulation (i.e. flatter diurnal cortisol slopes) within samples experiencing stress-related loss in GR sensitivity. This all lends support to the notion that over time, 84

chronic stress brings about dysregulation of the HPA axis and increased systemic inflammation via diminished sensitivity of the corticosteroid receptors. 2.10.2 Acute stress and corticosteroid receptor sensitivity Looking at the effects of acute stress on GR and MR sensitivity may shed some light on how stress-related loss of receptor sensitivity comes about. The effects of acute stress on corticosteroid receptor function have been examined in a number of studies. 2.10.2.1 Acute exercise stress and GR sensitivity Most of the early studies used exercise paradigms to examine acute stress-induced changes in receptor sensitivity to cortisol. DeRijk and colleagues examined the effects of dexamethasone on LPS-induced production of IL-6 in whole blood in healthy men exposed to graded exercise on a treadmill (DeRijk et al., 1996). Following exercise, more dexamethasone was required to inhibit the LPS-induced release of IL-6 indicating a reduction in GR sensitivity. The effects of dexamethasone on LPS-induced release of IL6, TNF-α, IL-10 and interferon (IFN)-γ in whole blood were examined in nine welltrained oarsmen who underwent strenuous exercise for a 15-20 minute period (Smits, Grünberg, Derijk, Sterk, & Hiemstra, 1998). Similar to the results of DeRijk and colleagues, following exercise, the inhibitory effect of dexamethasone on IL-6 and TNFα secretion was reduced indicating reduced GR sensitivity. However, dexamethasone effects on IL-10 and IFN-γ release were not altered by exercise. In contrast to the results of these studies, Duclos and colleagues looked at the effects of an acute bout of exercise on sensitivity to cortisol in the isolated cultured monocytes of endurance-trained men (n=6) and found an exercise-induced increase in GR sensitivity (Duclos et al., 1999). Similarly, in a more recent study, an acute resistance exercise

85

protocol in resistance-trained men and women (n=15) brought about increased GR expression in lymphocytes (Fragala et al., 2011). The reason for the discrepancy in results between exercise studies may be the use of different culture conditions across studies (Rohleder et al., 2003). As mentioned previously, performing glucocorticoid sensitivity assays in whole blood, as opposed to isolated PBMCs, does not take into account individual variability in white blood cell population ratios. 2.10.2.2 Acute psychosocial stress and GR sensitivity: Murine studies The effects of acute psychosocial stress on GR sensitivity have largely been investigated in animals. Sheridan and colleagues subjected mice to social reorganisation (SRO) and measured GR sensitivity using a synthetic glucocorticoid suppression test on proliferation of splenocytes (Sheridan, Stark, Avitsur, & Padgett, 2000). SRO stress involves randomly housing groups of male mice separately for two weeks in order for stable social hierarchies to form. The dominant mouse from each group is then transferred to a different cage where it is perceived as an aggressive intruder. This is stressful for both the resident mice and the intruder. The authors found that proliferation of splenocytes was inhibited in a dose-dependent manner by glucocorticoids in control mice, whereas proliferation of splenocytes in the SRO mice was resistant to glucocorticoid suppression. This indicates reduced GR sensitivity in the SRO mice brought about by acute psychosocial stress. Similarly, Stark and colleagues demonstrated that the splenocytes of SRO mice were resistant to the antiproliferative effects of corticosterone compared to control mice, suggesting a decrease in GR sensitivity following bouts of acute psychosocial stress (Stark et al., 2001). SRO exposure in mice has also been shown to downregulate the expression of GR mRNA (Quan et al., 2001). In all studies, resistance to glucocorticoids developed following repeat, but not acute, exposures to SRO, and the resistance persisted

86

for 10 days after the stress exposure ended (Avitsur, Stark, Dhabhar, Padgett, & Sheridan, 2002). Similarly, Bauer and colleagues showed that repeated exposure to restraint stress induced a slight increase in glucocorticoid resistance, i.e. decreased GR sensitivity (Bauer, Perks, Lightman, & Shanks, 2001). However, acute exposure did not induce any significant changes in GR sensitivity. 2.10.2.3 Acute psychosocial stress and corticosteroid receptor sensitivity: Human studies To date, five studies have assessed the effects of acute psychosocial stress on corticosteroid receptor sensitivity in humans. In these studies, participants were exposed to a number of behavioural tasks known to induce activation of the HPA axis stress response. In all studies (see Table 2.2) receptor sensitivity was measured using dexamethasone suppression of LPS-induced cytokine production in whole blood (see Section 6.8.2 for a more detailed description of this procedure). The first study measured sex differences in GR sensitivity following acute psychosocial stress in healthy young men and women (Rohleder, Schommer, Hellhammer, Engel, & Kirschbaum, 2001). Twenty-seven men and 18 women in the luteal phase of their menstrual cycle were exposed to the TSST (Kirschbaum, Pirke, & Hellhammer, 1993). Men and women did not differ in their salivary cortisol responses to acute stress. However, GR sensitivity showed marked gender differences. Examination of the inhibitory concentration 50% (IC50) of dexamethasone revealed that one hour after stress GR sensitivity had significantly increased in men, whereas sensitivity had decreased in women, although this change failed to achieve statistical significance. In agreement with these findings, the authors report that IL-6 levels one hour post-stress had significantly decreased in men but remained unchanged in women.

87

The second study measured age and sex-steroid related differences in GR sensitivity following acute psychosocial stress in healthy elder men (n=14), healthy young men (n=14), and healthy elder men who had received a testosterone injection five days prior to testing (n=12) (Rohleder, Kudielka, Hellhammer, Wolf, & Kirschbaum, 2002). All participants underwent the TSST. An hour after the stress protocol there were no differences between groups in terms of stress-induced increases in cortisol. However, GR sensitivity as indexed by the IC50 of dexamethasone was significantly increased in the younger men, and significantly decreased in the older men. Interestingly, testosteronetreated older men showed the same significant increase in GR sensitivity as the healthy younger men. These findings provide further evidence that acute stress modulates GR sensitivity. Furthermore, they indicate that GR sensitivity in response to stress changes with age and that these changes are associated with the presence of sex steroids. The third study examined the effects of oral contraception on GR sensitivity after acute psychosocial stress (Rohleder, Wolf, Piel, & Kirschbaum, 2003). Previous research has shown that women taking oral contraceptives have blunted cortisol responses to stress (Kirschbaum, Kudielka, Gaab, Schommer, & Hellhammer, 1999). HPA axis activation and GR sensitivity were measured in 14 women using oral contraception and 11 women in the luteal phase of the menstrual cycle that underwent the TSST. Following stress, luteal phase women showed an increase in cortisol whereas the contraceptive users showed blunted cortisol stress responses. Luteal phase women exhibited a non-significant decrease in GR sensitivity. Women taking oral contraceptives displayed an increase in GR sensitivity following acute stress. The authors posit that this increase in GR sensitivity is an adaptive response to the blunting of the cortisol stress reactivity which may protect women using oral contraceptives from the inflammatory stress response.

88

Table 2.2. Studies examining the effects of acute stress on corticosteroid receptor function Author/date

Rohleder et al. (2001)

Rohleder et al. (2002)

Rohleder et al. (2003)

Sample

45 healthy adults (18 women), mean age 25y

Study design

Stress paradigm

Difference between men and women in GR sensitivity following acute stress

TSST

40 healthy men, 14 young (mean age 25y), 14 elderly (mean age 67y), 12 elderly + testosterone treatment (mean age 68y)

Difference between young men, elderly men, and elderly men + testosterone in GR sensitivity following acute stress

TSST

25 healthy women , 14 taking OC (mean age 22y), 11 OC-free (mean age 25y)

Difference between women taking OC, and women not, in GR sensitivity following acute stress

TSST

GR/MR measurement protocol

Statistical test and covariates

DEX suppression of LPSinduced IL-6 and TNF-α production in whole blood

ANOVA; no covariates

DEX suppression of LPSinduced IL-6 and TNF-α production in whole blood

ANOVA; no covariates

Main findings

Basal GR sensitivity lower in men. Increase in GR sensitivity in men, and a non-sig decrease in women 60 mins after acute stress Basal GR sensitivity lower in younger men. Increase in GR sensitivity in young and testosterone-treated elderly men, non-sig decrease in elderly men, 60 mins after acute stress.

DEX suppression of LPSinduced IL-6 and TNF-α production in whole blood

ANOVA; no covariates

No difference in basal GR sensitivity Increase of GR sensitivity in OC users, no sig. change in women not taking OC.

DEX = dexamethasone; GLM= general linear model; GR = glucocorticoid receptor; IL-6 = interleukin-6; LPS = lipopolysaccharide; MR = mineralocorticoid receptor; OC = oral contraception; PRED = prednisolone; TNF-α = tumour necrosis factor – α; TSST = Trier Social Stress Test.

89

Table 2.2. (Continued) Studies examining the effects of acute stress on corticosteroid receptor function Wirtz et al. (2008)

Carvalho et al. (2015)

42 healthy men (mean age 43y)

74 older adults, 37 with T2DM (mean age 64y), 32 healthy controls (mean age 67y)

Association between BMI and GR sensitivity following acute stress

TSST

Difference between adults with T2DM and healthy controls in GR and MR sensitivity following acute stress

2x 5 min behavioural tasks

DEX suppression of LPSinduced TNF-α production in whole blood

DEX and PRED suppression of LPSinduced IL-6 production in whole blood

ANCOVA, GLM; baseline GR sensitivity, age, mean arterial pressure

Basal GR sensitivity not associated with BMI.

GLM; BMI, time of session

T2DM group had higher GR sensitivity at baseline, but not MR sensitivity.

Higher BMI associated with decrease in GR sensitivity after acute stress

Decrease in GR and MR sensitivity in healthy controls, no change on T2DM

DEX = dexamethasone; GLM= general linear model; GR = glucocorticoid receptor; IL-6 = interleukin-6; LPS = lipopolysaccharide; MR = mineralocorticoid receptor; OC = oral contraception; PRED = prednisolone; T2DM = type 2 diabetes; TNF-α = tumour necrosis factor – α; TSST = Trier Social Stress Test.

90

The fourth study investigated whether BMI affected changes in GR sensitivity following acute psychosocial stress (Wirtz, Ehlert, Emini, & Suter, 2008). Forty-two men underwent the TSST. BMI was not associated with either diurnal or stress-induced cortisol secretion. However, results indicated that a higher BMI was associated with a more pronounced loss of GR sensitivity following acute stress. The authors suggest that this could be a pathway through which BMI might alter the stress response in ways that are detrimental to cardiovascular health. The fifth study carried out by our group examined both GR and MR sensitivity to acute stress in 37 people with type 2 diabetes and 37 healthy controls (Carvalho et al., 2015). People with type 2 diabetes have an increased risk of CVD as well as impairments of the HPA axis (Bruehl et al., 2007; Hackett, Steptoe, & Kumari, 2014). MR sensitivity was measured using prednisolone suppression of LPS-induced cytokine production in whole blood (see Section 6.8.2 for a full description of the procedure). Prednisolone is a synthetic glucocorticoid that binds to both the GR and the MR. Mental stress was induced using two 5-minute behavioural tasks. Following stress, the healthy controls (mean age = 67.5 years) exhibited a decrease in both GR and MR sensitivity, which is in line with the previous finding that GR sensitivity decreases in healthy older men (Rohleder et al., 2002). However, there was no change in GR or MR sensitivity in those with type 2 diabetes. The diabetic patients also had blunted stress responses in terms of systolic blood pressure, heart rate, and levels of IL-6. The authors suggest that the impaired stress responsivity in type 2 diabetes is in part due to a lack of stress-induced alterations in GR and MR sensitivity. Apart from this fifth study, very little work has been done examining the effects of acute stress on the MR. Studies have shown that MR antagonists such as spironolactone result in increased basal cortisol levels and increased cortisol responses to exercise stress 91

(Heuser, Deuschle, Weber, Stalla, & Holsboer, 2000; Wellhoener, Born, Fehm, & Dodt, 2004). A common polymorphism in the MR gene has been associated with higher cortisol responses to acute stress (DeRijk et al., 2006). As the MR and GR work in concert to regulate the cortisol and inflammatory stress response, future stress research should examine both GR and MR sensitivity in order to gain further understanding of the link between stress and CVD. Therefore, the third study of this PhD presented in Chapter 6 will examine the effects of an acute psychosocial stress paradigm on both GR and MR sensitivity in healthy volunteers. To summarise, results from studies examining the effects of acute stress on corticosteroid receptor sensitivity have been mixed. Murine studies suggest that acute stress brings about a decrease in GR sensitivity. However, it could be argued that these studies adopt a sub-chronic stress paradigm as the effects on GR sensitivity are only seen after repeated exposures to the stressor. Human studies have provided varied data on the effects of both exercise and psychosocial stress on GR sensitivity. The main conclusion that can be drawn from results so far is that acute stress modulates GR sensitivity. There is rather large variability in corticosteroid receptor sensitivity in humans with regards to sex, age, sex steroid hormone status, BMI, as well as diabetes status. 2.11 Chapter summary Although the direction of results is mixed, psychosocial stress factors and negative emotional disorders appear to be associated with dysregulation of both basal and stressrelated HPA axis function. Dysregulation of both basal and stress-related HPA axis function has been associated with markers of cardiovascular risk and have been seen in CVD patients. This evidence suggests that dysregulation of the HPA axis is likely one of 92

the biological pathways through which psychosocial stress contributes to the pathophysiology of CVD. The evidence also suggests that alterations in the sensitivity of the corticosteroid receptors may be one of the mechanisms through which psychosocial stress brings about sustained changes in HPA axis function. The studies cited in the previous sections provide support for the notion that stress modulates corticosteroid sensitivity. Reduced GR sensitivity has been reported in depression. Chronic life stressors, such as job stress and caregiver stress, have been shown to reduce GR sensitivity also. Repeated stress ‘hits’ over time may result in a loss of receptor sensitivity, thereby leading to dysregulated cortisol secretion, and increased systemic inflammation. For example, in CHD patients, 24-hour cortisol secretion is higher than healthy controls and this is accompanied by higher levels of CRP and IL-6 (Nijm et al., 2007). This implies diminished corticosteroid receptor sensitivity in these patients. Data from studies assessing the effects of acute stress on corticosteroid receptor sensitivity are more mixed. There is a large amount of variability in GR sensitivity following stress in humans with regards to sex, age, and BMI. Moreover, work examining the effects of acute stress on MR sensitivity is scarce. Nevertheless, results of these studies show that acute stress does modulate corticosteroid receptor sensitivity. Taken together, the evidence suggests that dysregulation of the HPA axis, via stress-related modulation of the corticosteroid receptors, is one of the biological pathways linking psychosocial stress and CVD.

93

Chapter 3 Study 1 - Diurnal cortisol rhythm and adverse clinical outcomes in patients with advanced CVD: The ARCS Study 3.1 The Adjustment and Recovery after Cardiac Surgery (ARCS) Study Coronary artery bypass graft (CABG) surgery is used to relieve symptoms and improve life-expectancy in those suffering from advanced coronary heart disease. The ARCS Study was designed to investigate the causes and consequences of poor physical and emotional wellbeing following CABG surgery, and the implications for patient quality of life and physical recovery. Five sets of factors potentially relevant to emotional and physical quality of life post-CABG surgery were the focus of the study: (1) clinical factors, e.g. existing heart problems and illness as well as factors pertaining to the surgery itself, (2) cognitive factors, e.g. cognitive function as well as the patients’ ability to understand health information, (3) social factors, e.g. social support, (4) emotional factors, e.g. depression and anxiety, (5) biological factors, e.g. inflammatory markers measured in the blood and salivary cortisol measured across the day. The ARCS Study used a prospective longitudinal design with a number of assessment periods spanning up to 2.68 years after the CABG procedure. Patients were recruited at their surgical pre-assessment clinic and were assessed approximately one month prior to their surgery (T1), 4-5 days after their surgery while still in hospital care (T2), 8-10 weeks after surgery (T3), and 12 months after surgery (T4). At each time point, participants were asked to complete a questionnaire pack and provide saliva samples across the day for measurement of diurnal cortisol profiles (saliva was not provided at the visit 4-5 days after surgery). Blood measures were taken prior to surgery and in the days following surgery in order to measure markers of inflammation. Approximately 2.5 years following 94

the procedure, long term clinical outcomes for each patient were collected from electronic and paper medical records (T5). This included mortality data, development of postsurgical infections, any cardiac or non-cardiac related hospital readmissions, adverse cardiac events, occurrence of other cardiac procedures or tests (e.g. angiogram, percutaneous coronary intervention), occurrence of new onset depression or anxiety, and diagnoses of any other major illnesses. 3.2 My contribution to the ARCS Study As part of a team of several ARCS Study researchers, I was involved in study recruitment and data collection at all time-points. I recruited a large number of patients at their surgical pre-assessment. As well as explaining the study to the patient and obtaining informed consent, this also involved administering a short cognitive examination and health literacy test, as well as organising blood sample collections for each patient. In terms of data collection, I carried out a large number of on-ward structured interviews with patients approximately 4-5 days after surgery. I also sent questionnaire and salivacollection packs to patients at the 8-10 week and 12 month follow-up points. Additionally, I was responsible for prompting patients over the telephone who may have forgotten to return their questionnaire packs in the post. My largest contribution to the ARCS Study was the collection of the long-term clinical outcomes which I was responsible for. In the early stages of my PhD, I spent a number of months on site at St. George’s hospital collecting long-term clinical outcome data for each individual patient from electronic and paper medical records. Additionally, I was largely involved in ARCS Study data entry as well as maintenance of the dataset. Furthermore, I have been involved in data analysis. To date, I have produced two first-author publications using ARCS data, and have contributed to several other 95

ARCS Study publications. (Kidd et al., 2014; Kidd, Poole, Leigh, et al., 2016; Kidd, Poole, Ronaldson, et al., 2016; Poole et al., 2015; Poole, Kidd, et al., 2014a, 2014b, 2016; Poole, Leigh, et al., 2014; Poole, Ronaldson, et al., 2016; Ronaldson et al., 2014, 2015; Steptoe et al., 2015). 3.3 Differentiating my PhD from the ARCS Study The ARCS Study is a multidisciplinary study involving several researchers. This study has produced a rich dataset containing information pertaining to the five sets of factors outlined previously. Accordingly, many issues have been and will be investigated that are beyond the scope of my PhD. In my PhD, I used pre-surgical data from T1 of the ARCS Study to examine the association between pre-surgical diurnal cortisol rhythm and major adverse cardiac events (MACE) and death (T5 data) in patients with advanced heart disease undergoing CABG surgery. I also used T1 data to cross-sectionally explore what psychosocial stress factors may be affecting diurnal HPA axis function. Results from this study have been published in the Journal of Clinical Endocrinology & Metabolism (Ronaldson et al., 2015). 3.4 Introduction As mentioned previously in this thesis, there is growing evidence that the HPA axis plays a role in the progression of CVD. Elevated 24h urinary cortisol has been found to predict cardiovascular death in older people both with and without CVD (Vogelzangs et al., 2010). Higher serum cortisol has also been associated with cardiovascular mortality in a cohort of patients with mood disorder (Jokinen & Nordström, 2009). However, the role of the HPA axis in patients with advanced CVD is less clear. Higher serum cortisol levels have been found to predict both mortality risk and risk of future cardiac events in chronic heart failure (Güder et al., 2007; Yamaji et al., 2009) and ischaemic stroke (Barugh, Gray, 96

Shenkin, MacLullich, & Mead, 2014). However, results from studies of cortisol in acute coronary syndrome have been less consistent (Jutla, Yuyun, Quinn, & Ng, 2014; Reynolds et al., 2010). One difficulty in interpreting this evidence is that cortisol is typically measured with a single serum sample. Inconsistencies in associations between cortisol and CVD may be because the diurnal nature of cortisol is not being taken into account. More detailed measurement of the diurnal cortisol profile would allow for a more in depth investigation of the associations between cortisol and clinical endpoints in CVD patients. Dysregulation of the HPA axis can result in a reduction in the amplitude of the diurnal pattern, or a flatter slope across the day. As mentioned earlier in Chapter 2, flatter cortisol slopes have been associated with higher levels of coronary artery calcification (Hajat et al., 2013; Matthews et al., 2006), and increased cardiovascular mortality in nonclinical populations (Kumari et al., 2011). There is a paucity of studies examining the effects of variations in diurnal cortisol rhythms on future cardiac events and mortality in patients with established CVD. This study therefore sought to examine the relationship between pre-surgical diurnal cortisol and clinical outcomes in patients undergoing CABG surgery. 3.4.1 Hypotheses Based on previous research, I hypothesised that a flatter diurnal cortisol slope before surgery would be associated with higher rates of future cardiac events and mortality in the years following CABG. I also examined associations between the cortisol awakening response (CAR) and total cortisol output across the day, and adverse clinical outcomes. However, in keeping with previous research I did not expect to find significant associations with these cortisol parameters (Kumari et al., 2011; Matthews et al., 2006). 97

As a flatter diurnal slope could reflect a negative psychosocial stress profile I carried out exploratory analyses examining cross-sectional associations between pre-surgical cortisol slopes and psychosocial stress variables, namely stressful life events, depression, anxiety, and social support, in order to garner information about stress-related factors that may bring about dysregulation of the HPA axis. I hypothesised that flatter cortisol slopes would be associated with more depressive symptoms, higher levels of anxiety, more stressful life events, and low social support. 3.5 Materials and methods 3.5.1 Participants The data we used in this analysis were collected as part of the ARCS Study, involving patients undergoing first-time elective CABG surgery or CABG plus valve replacement. CABG surgery in a single centre (Steptoe et al., 2015) included both on-pump and offpump procedures. All procedures were carried out with written informed consent of the participants. Ethical approval was obtained from the National Research Ethics Service. Participants were 262 prospective CABG patients who were recruited from a pre-surgical assessment clinic at St. George’s Hospital, London. Eligible participants had to be at least 18 years of age and had to be able to complete questionnaires in English. Long term recovery outcomes were collected from electronic and paper patient records on average 2.68 years (SD = 0.40) after surgery. We carried out analyses on 250 patients with complete data on clinical outcomes and cortisol slope. There were no significant associations between the use of steroid medications and cortisol output, outcome variables or covariates (all p values > 0.05). Therefore patients taking steroid medications (n = 8) were included in the analyses.

98

There were no significant differences between patients included in and excluded from the analyses in terms of age, sex, BMI, smoking status, length of hospital stay, the occurrence of major adverse cardiac events (MACE), chronic disease burden, diabetes status, or whether or not the person had on-pump surgery (all p values < 0.05). However, European System for Cardiac Operative Risk Evaluation (EuroSCORE) was higher in the 12 patients without cortisol data (F(2, 345) = 5.23, p = 0.006) indicating poorer prognosis on average. Patients included in the analyses did not differ from those excluded in terms of any of the psychosocial stress variables (all p values < 0.05). 3.5.2 Biological and clinical measures Diurnal salivary cortisol At the pre-surgical assessment clinic (T1) participants received a saliva collection kit and were given instructions for collection at home. The kit included seven pre-labelled ‘salivette’ collection tubes (Sarstedt, Leicester, UK) and a cortisol diary. The cortisol diary contained instructions on how and when to give samples (Appendix A). These diaries were also used to record information on factors likely to introduce variation in cortisol samples such as mood, exercise, and daily stressors. Participants were instructed to choose one day prior to surgery on which to provide seven saliva samples at set time points: on waking, 30 minutes after waking (30+), 10am, 12pm, 4pm, 8pm, and bedtime. Participants stored their samples in the refrigerator before returning them to the clinic. The samples were obtained an average 30.6 days (SD = 36.9) prior to surgery and were stored at -20°C for analysis at a later date. Cortisol levels were assessed from saliva using a time resolved immunoassay with fluorescence detection at the University of Dresden, Germany. The intra- and inter-assay coefficients of variation were less than 4%.

99

We computed three different cortisol measures. Total cortisol output over the day was assessed by calculating the cortisol AUC with respect to ground (Pruessner, Kirschbaum, Meinlschmid, & Hellhammer, 2003). The CAR was calculated by measuring the difference between the sample taken on waking and the 30+ sample. In line with other work produced by Steptoe’s group, participants who reported giving their first sample more than 15 minutes after waking were excluded from the analyses. Previous research has shown that a long delay between waking and providing the ‘waking’ sample can produce misleading CAR results, but a delay of less than 15 minutes between waking and providing the sample does not seem problematic (Dockray, Bhattacharyya, Molloy, & Steptoe, 2008). An expert panel recently recommended that CAR data should be excluded if the waking sample is provided with a delay of 5 minutes or more. However, the same expert panel also stated that this tight accuracy margin would result in substantial data loss (26-46%) and therefore researchers need to choose between scientific precision and practical feasibility (Stalder et al., 2016). A number of previous studies have selected an accuracy margin of 0.05). 169

5.8 Discussion 5.8.1 Summary of results The aim of this study was to assess the effects of six-day administration of beta-blockers and SSRIs on several different indices of diurnal HPA axis function. We hypothesised that both drugs would bring about changes in diurnal cortisol secretion. More specifically, we hypothesised that beta-blockers would increase diurnal cortisol secretion, in that the CAR and cortisol AUC would be enhanced leading to flatter cortisol slopes. We hypothesised that SSRIs would lead to reduced cortisol output with lower waking cortisol levels, a reduced CAR and cortisol AUC, and steeper cortisol slopes. We also postulated that sex would play a role in the effects of the study drugs on cortisol secretion over the day. The results of this study provide limited support for these hypotheses. There were no effects of beta-blockade on cortisol dynamics. Compared with placebo, women taking SSRIs had significantly steeper cortisol slopes across the day. This observed difference in cortisol slope was independent of any differences in stress- or mood-related factors, suggesting that the observed results were due to direct biological effects of SSRIs on HPA axis function. The group taking SSRIs did not differ significantly on any other cortisol parameter. 5.8.2 Comparison to previous research Our results are in line with those of Rota et al. (2005) who found that 14 days administration of fluvoxamine to 20 depressed individuals resulted in a steepening of the cortisol slope. However, our findings are not in line with those of Hibel and colleagues and Tucker and colleagues who found that SSRIs had no significant effect on cortisol rhythm across the day in children with problem behaviour and depressed patients respectively (Hibel et al., 2007; Tucker et al., 2004). 170

It is difficult to compare the results of our study to others. This is largely because previous research has assessed effects of SSRIs on HPA axis function in clinical samples. As outlined previously, long-term treatment with SSRIs has been shown to reduce the cortisol AUC, as well as reduce waking levels of cortisol in patient samples with depression and generalised anxiety disorder (Hinkelmann et al., 2012; Knorr et al., 2012; Lenze et al., 2011; Ruhé et al., 2015). Cortisol AUC is known to be altered in depression and anxiety (Heaney et al., 2010; Marchand et al., 2014). Therefore, the effects of SSRI treatment on cortisol AUC may be more pronounced in those who have these stressrelated illnesses. We examined the effects of SSRIs in healthy volunteers which may explain why we did not observe any significant effects on overall daily cortisol output. In the current study, we found that women taking escitalopram had higher waking levels of cortisol compared to women taking placebo. This alteration in morning cortisol likely drove the significant changes in cortisol slope in this group. This finding is in line with that of Harmer and colleagues who found that six days administration of citalopram (20mg/day) brought about significant increases in waking cortisol in healthy volunteers (Harmer, Bhagwagar, Shelley, & Cowen, 2003). Conversely, in depressed patients, SSRIs have been found to lower levels of waking cortisol (Knorr et al., 2012; Ruhé et al., 2015). Waking levels of cortisol have been found to be increased in depression (Bhagwagar et al., 2005; Mannie et al., 2007). Therefore, the direction of the effect of SSRIs on waking cortisol may be related to mental health status – in depressed patients SSRIs ‘normalise’ elevated levels of waking cortisol, and in healthy individuals SSRIs increase waking levels. More research is needed to confirm this. It is possible that SSRI dosage (10mg/day) used in the current study was not sufficient to elicit changes in certain diurnal cortisol parameters in healthy volunteers (i.e. cortisol AUC, CAR, evening cortisol levels). Previous studies assessing the effects of 171

escitalopram initially prescribed 10mg per day to participants but then titrated up to 20mg as the study progressed (Hinkelmann et al., 2012; Park et al., 2015). Up-titration was not practicable in the current study due to the study duration. Perhaps if we had increased the dosage we would have observed significant effects of escitalopram on other diurnal cortisol parameter also. Additionally, we may have observed significant effects if we increased the duration of treatment. Previous studies have shown that 12 weeks treatment with paroxetine and escitalopram have significantly reduced the CAR in patients with major depressive disorder and generalised anxiety disorder respectively (Lenze et al., 2011; Ruhé et al., 2015). In this study SSRIs had no effect on the CAR. This may be for reasons to do with using a healthy sample as the CAR has been shown to be altered in depression and anxiety (Dedovic & Ngiam, 2015; Veen et al., 2011; Wardenaar et al., 2011), or it may be a power issue. As mentioned in Chapter 3, the accuracy margin of 50 years) having poorer responses to SSRIs (Thase, Entsuah, Cantillon, & Kornstein, 2005). The current study was carried out in a relatively young sample, meaning that the results may not be 175

replicable in older patient sample, like in those with CVD. This area warrants further investigation, and future studies should take age and sex differences into consideration. 5.8.6 Beta-blockers: Lack of effect In the current study, beta-blockers had no significant effect on any of the diurnal cortisol parameters. This is in line with previous findings where long-term administration of betablockers brought about no changes in plasma cortisol levels, although these studies did not assess diurnal cortisol patterns (Golub et al., 1981; Rosen et al., 1988). As mentioned in Chapter 2, the SAM system and the HPA axis are the major biological systems involved in the stress response, and these systems interact during times of stress. There is also evidence to suggest that these two systems interact under basal, unstressed conditions. In both healthy and depressed patients diurnal variations cerebrospinal fluid levels of norepinephrine and plasma cortisol levels are very highly correlated (Wong et al., 2000). Therefore it is puzzling that under unstressed conditions beta-blockade did not bring about alterations in diurnal cortisol secretion in this study. It is possible that the propranolol dosage used in the current study was not sufficient to elicit changes in diurnal cortisol parameters in healthy volunteers. An 80mg dose failed to bring about changes in the current study. However, 80mg of propranolol was found to be sufficient to induce change in basal plasma levels of night-time cortisol in healthy men (Dart et al., 1981). These alterations in night-time cortisol levels were observed after sixweek administration of propranolol. It is possible that the treatment duration of the current study (six days) was not sufficient to elicit changes in indices of diurnal HPA axis function. Although, administration of acute doses of propranolol (ranging from 10-80mg) have brought about increases in circulating cortisol levels in healthy and diabetic volunteers (Kizildere et al., 2003; Lewis et al., 1981; Popp et al., 1984). Propranolol is a

176

rapidly metabolised drug that exerts effects on the β-adrenergic receptors quickly (Leahey et al., 1980). Propranolol has a half-life of about three to four hours and there is large variability in bioavailability across individuals due to rapid metabolism of propranolol in the liver (Gomeni, Bianchetti, Sega, & Morselli, 1977). In the current study, participants were instructed to take propranolol every morning after their breakfast. However, there is no guarantee participants took the drug at the same time every day, or with food which is known to affect its bioavailability (Liedholm, Wåhlin-Boll, & Melander, 1990). Therefore, although participants received sustained-release propranolol, it is possible that there may have been a high rate of variability between participants in the current study regarding the bioavailability of propranolol. This may be why there were no significant effects on diurnal HPA axis function. Future studies should determine blood concentrations of propranolol at the end of the treatment period in order to adjust for this factor. As mentioned in Chapter 4, propranolol was chosen to assess the effects of changes in the peripheral nervous system on HPA axis function. However, as well as exerting effects on the beta-adrenergic receptors, there is evidence that propranolol also might act as an antagonist of the serotonin receptors 5-HT1A and 5-HT1B (Davids & Lesch, 1996; Hoyer et al., 1994). Therefore, propranolol may not have been the most appropriate pharmacological probe for assessing the effects of sympathetic activation on HPA axis function. Future research should seek to use beta-blockers that do not act on serotonin receptors, such as the selective beta-blocker metoprolol. 5.8.7 Strengths and limitations A strength of this study is that it was a randomised placebo-controlled double-blind trial. We adopted a parallel group design meaning that participants receiving placebo acted as

177

the control comparison group for both experimental medications. The three groups in the study did not differ significantly in demographic or stress-related factors. As we did not employ a crossover design it is possible that the treatment groups were unbalanced on some covariates that were not measured in the study. However, adopting a parallel groups design allowed us to avoid problems relating to order and carry-over effects to do with the study medications. It also meant that participants were unable to become habituated to the stress protocol. This study had a retention rate of 88.5% with 94 participants providing usable data on some parameters of diurnal cortisol secretion. However, it is possible that this study was underpowered to detect certain effects. There were also more women than men in the current study, meaning that we may have lacked sufficient statistical power to detect drug effects in men. Additionally, our sample was largely composed of healthy university students from high socioeconomic backgrounds. Therefore the results may not be readily generalizable to other groups, or to clinical groups with depression or CVD. Cortisol was measured over a single day meaning that the diurnal secretion may have been affected by situational factors rather than long-term factors. As mentioned previously in Chapter 3 of this PhD, diurnal indices of HPA axis function are predominantly affected by trait factors on weekdays as most people have established weekday routines (Hellhammer et al., 2007). In the Stress Pathways Study we measured cortisol over the course of a weekday which may help counteract the effects of single-day sampling. However, the majority of participants were students meaning that routine may have been quite variable across weekdays. Therefore the diurnal cortisol profiles of participants may have reflected state-like properties rather than trait-like influences. This measurement issue should be borne in mind while interpreting results.

178

One further limitation of this study was the use of multiple testing. Within each comparison group the effects of experimental condition on seven different cortisol parameters were measured simultaneously. This means that the probability of observing a significant result due to chance was increased. Use of the Bonferroni correction would have set the significance cut-off at p < 0.007, thereby rendering the effects of SSRIs on cortisol in females non-significant. However, the Bonferroni correction has a tendency to be too conservative (Narum, 2006), and the replication of the significant effects of SSRIs in both cortisol slope and wake/bedtime difference in women implies that this result was not down to chance. 5.8.8 Conclusion In conclusion, the results of this study indicate that six day treatment with the SSRI escitalopram may bring about a steepening of cortisol slopes in healthy women, via increases in waking cortisol levels. Flattened cortisol rhythms have been associated with chronic stress, depression, and CHD. This finding suggests that flattening of the cortisol slope in women may be related to alterations in the serotonergic system. It also implies that SSRIs may exert their therapeutic effects in women via correction of a flattened diurnal cortisol rhythm. However, due to the various methodological limitations of this study future research is needed to replicate this result. In the following chapter I will examine the effects of these study medications on cortisol stress reactivity, corticosteroid receptor sensitivity, and other stress-related factors which may help to further explain the results of the current study.

179

Chapter 6 Study 3 – The Stress Pathways Study results: The effect of pharmacological blockade on cortisol stress reactivity and corticosteroid receptor sensitivity in healthy volunteers 6.1 Introduction In this chapter I will present results from the Stress Pathways Study concerning the effects of beta-blockade and SSRIs on cortisol secretion in response to acute psychosocial stress. I will also present results regarding the effects of acute psychosocial stress on corticosteroid receptor sensitivity. Moreover, I will examine the effects of the study medications on baseline corticosteroid receptor sensitivity, and changes in receptor sensitivity after acute stress. In order to paint a broader picture of how the study medications are affecting the biological stress response, I will examine the effects of the medications on cardiovascular stress reactivity. Together with the results from Chapter 5 of this thesis, these results may shed light on the biological mechanisms involved in dysregulation of HPA axis, and highlight medications that might be suitable for the treatment of impaired HPA axis function. 6.2 Literature review: Beta-blockers and cortisol stress reactivity In Chapter 5 I described how the HPA axis and the SAM system are anatomically linked and how they interact with each other in order to regulate a number of stress-related functions, including the release of pro-inflammatory cytokines. However, under basal, unstressed conditions I found that six-day treatment with propranolol brought about no significant changes in diurnal HPA axis function. It is possible that propranolol may affect 180

stress-related cortisol levels. A number of studies have investigated the effects of betablockade on cortisol responses to a number of types of stress which will be outlined below. 6.2.1 Acute administration of beta-blockers Early investigations in to the effects of beta-blockade on stress-related HPA axis function examined cortisol responses to exercise. MacDonald and colleagues administered either a single dose of metoprolol (100mg), propranolol (80mg), or placebo in a crossover design to 10 men with essential hypertension (Macdonald, Bennett, Brown, Wilcox, & Skene, 1984). The men then underwent a prolonged exercise protocol. The single dose of propranolol, but not metoprolol, brought about significant increases in plasma cortisol and adrenaline levels during exercise compared to placebo. Similar increases in cortisol responses to submaximal exercise have been observed in healthy untrained men receiving acute intravenous pre-treatment with propranolol (Jezová, Vigas, Klimes, & Jurcovicová, 1983). In agreement with these earlier studies, a more recent study reported increases in plasma cortisol levels in ten healthy men undergoing maximal exercise to exhaustion after a dose of 80mg propranolol compared to placebo (Viru et al., 2007). However, single doses of 150mg metoprolol and 120mg propranolol brought about no significant changes in cortisol secretion following exercise (30m cycle) in seven healthy men compared to placebo (Uusitupa et al., 1982). Acute effects of beta-blockers on cortisol reactivity to psychosocial stress have also been examined. Andrews and Pruessner, in what they called the ‘propranolol suppression test’ administered a single dose of propranolol (80mg) or placebo to 30 healthy men (n=15 in each group) (Andrews & Pruessner, 2013). Following this, participants underwent the TSST. Results showed that those who received propranolol had significantly increased

181

cortisol responses to acute stress alongside significantly decreased heart rate responses, compared to the placebo group. A single 80mg dose of propranolol given to 14 healthy men who underwent a psychosocial stress protocol resulted in higher salivary cortisol stress responses compared to placebo-treated men (Maheu, Joober, & Lupien, 2005). In six young Type A men, a single dose of propranolol attenuated heart rate responses and increased cortisol responses to mental arithmetic stress (Williams, Lane, Kuhn, Knopes, & Schanberg, 1988). However, Dreifus and colleagues administered 60mg propranolol or placebo to 48 healthy women prior to undergoing the TSST and found that both groups experienced stress-induced cortisol increases that did not differ significantly (Dreifus, Engler, & Kissler, 2014). It is possible that the 60mg dose administered by Dreifus and colleagues (2014) was not adequate to bring about changes in stress-related HPA axis function, seeing as the majority of studies reporting cortisol increases administered 80mg propranolol. More novel stress paradigms have been used to investigate the effects of acute betablockade on stress-related HPA axis function. Benschop and colleagues report that 160mg (4x40mg doses) propranolol administered over 1.5 days to 16 healthy men undergoing their first-time parachute jump elicited no effects in cortisol stress reactivity compared to placebo (Benschop et al., 1996). However, the men receiving propranolol did have more pronounced ACTH responses to the jump than those receiving placebo (Oberbeck et al., 1998). Cortisol responses to cold water immersion stress were also increased by a 40mg dose of propranolol in eight healthy young men (Šimečková, Janskỳ, Lesna, Vybiral, & Šrámek, 2000). Khan and colleagues examined the effects of an acute intravenous dose of propranolol (0.2mg/kg) on cortisol responses to pentagastrin administration in 16 healthy men and women (Khan, Liberzon, & Abelson, 2004). Pentagastrin is a substance that produces symptoms of anxiety and panic, and brings about strong activation of the 182

HPA axis. Compared to placebo, those who received propranolol had a delayed but enhanced cortisol response to pentagastrin. ACTH and adrenaline responses were also enhanced in the propranolol group. The heart rate acceleration usually brought about by pentagastrin was virtually eliminated by propranolol also. Together, what these results suggest is that acute beta-blockade brings about enhanced cortisol stress reactivity. At the same time, beta-blockade also attenuates heart rate responses to stress. This implies an inverse relationship between the HPA axis and the SAM system in that suppression of the SNS stress response by beta-blockade brings about an increase in HPA axis function. In support of this, a study that combined the DST with the TSST found that those who had received dexamethasone the night before stress testing had lower cortisol stress responses in combination with an elevated heart rate compared to the placebo group (Andrews, D’Aguiar, & Pruessner, 2012). 6.2.2 Long-term administration of beta-blockers The effects of more long-term administration of beta-blockers on cortisol stress reactivity have also been examined, but to a lesser extent than acute administration. Two early studies examined the effects of long-term beta-blockade on cortisol responses to exercise. In the first study, 10 men with essential hypertension received 28 day treatment with propranolol (80mg/day), metoprolol (100mg/day), and placebo in a crossover design (Macdonald et al., 1984). Following this, the men underwent a prolonged exercise protocol. Both the propranolol and metoprolol treatment brought about increased cortisol responses to exercise compared to placebo. Similarly, in 18 healthy young men 100mg metoprolol (twice daily) and 10mg timolol (twice daily) for five days resulted in significantly increased cortisol responses to exercise compared to placebo (Gullestad, Dolva, Kjeldsen, Eide, & Kjekshus, 1989).

183

To date, only one study has examined the effects of prolonged beta-blockade on cortisol responses to psychosocial stress. Kudielka and colleagues administered propranolol (80mg/day) to 19 healthy men and women for five days (Kudielka et al., 2007). Participants then underwent the TSST. Neither pre-stress cortisol levels nor cortisol responses to the TSST in the propranolol group differed significantly from placebo. Similar null findings have been reported in 20 healthy male volunteers undergoing a bungee jump who received three day pre-treatment with propranolol (3 x 40mg/day) (van Westerloo et al., 2011). 6.2.3 Summary Taken together, the studies examining effects of acute beta-blockade on cortisol stress reactivity suggest that acute suppression of the SNS via beta-blockade seems to enhance HPA axis stress reactivity. This effect appears to hold despite heterogeneity between studies in terms of sample size, stress paradigm, route of drug administration, and biological specimen used for cortisol measurement (saliva/plasma). However, results from studies examining more long-term effects of beta-blockade on cortisol stress reactivity have been mixed. Five-day and 28-day administration of beta-blockers appear to increase cortisol secretion following exercise paradigms. However, using a psychosocial stress paradigm and a more novel bungee jump paradigm has produced null findings. This is peculiar considering acute doses appear to bring about enhanced cortisol responses to the TSST. Interestingly, the studies that report these enhanced responses do so only in male samples (Andrews & Pruessner, 2013; Maheu et al., 2005; Williams et al., 1988). Dreifus and colleagues reported null findings following the TSST in a sample comprised of women (Dreifus et al., 2014). The inclusion of women in the sample may have been one of the reasons Kudielka and colleagues (2007) report null findings following longer-term administration of propranolol. Neither study include the use of 184

contraceptive pill as a covariate – something that is known to affect cortisol stress reactivity (Rohleder, Wolf, Piel, & Kirschbaum, 2003). This may also have affected results obtained. In sum, the results of this body of work suggest that suppression of the SNS brings about increases in stress-related HPA axis function. However, further work is needed examining effects of longer-term beta-blocker administration. 6.3 Literature review: SSRIs and cortisol stress reactivity In Chapter 5 I reported results from the Stress Pathways Study showing that SSRIs brought about steeper diurnal cortisol rhythms in women. Flatter slopes are known to be characteristic of depression (Doane et al., 2013; Jarcho et al., 2013; Sjögren et al., 2006) and the results from the Stress Pathways Study suggest that SSRI treatment may normalise the diurnal cortisol rhythm in women. Cortisol stress reactivity is also known to be dysregulated in depression with both enhanced and blunted cortisol responses to stress being reported depending on the severity and duration of depressive symptoms. It is possible that SSRIs may affect the cortisol stress response also. A number of studies have examined these effects. 6.3.1 Acute administration of SSRIs Using a crossover design, Ahrens and colleagues administered single doses of either the SSRI sertraline (100mg) or placebo to 12 healthy men (Ahrens, Frankhauser, Lederbogen, & Deuschle, 2007). They then examined neuroendocrine responses to exercise stress. Baseline pre-exercise cortisol levels in the sertraline group were higher, and the cortisol responses to exercise stress were enhanced in this group also. The effects of acute SSRI administration on cortisol response to psychosocial stress have also been examined.

185

Healthy men were randomised to receive single doses of 10mg escitalopram (n=17), 20mg escitalopram (n=14), or placebo (n=12) (Garcia-Leal, Del-Ben, Leal, Graeff, & Guimarães, 2010). Following this, they underwent a simulated public speaking protocol, similar to the public speaking component of the TSST. Escitalopram did not bring about any significant alterations in cortisol stress reactivity compared to placebo. However, this is probably because the public speaking protocol did not elicit a cortisol stress response in either group. 6.3.2 Long-term administration of SSRIs A number of studies have examined the effects of longer-term SSRI administration on cortisol stress reactivity, with mixed results. Ljung and colleagues administered six months treatment with citalopram (20-40mg/day) or placebo to 16 healthy men with moderate abdominal obesity in a crossover trial (Ljung et al., 2001). Following the treatment period, the men underwent an arithmetic stress test. Citalopram brought about increases in baseline morning cortisol values, and following stress cortisol secretion was enhanced in the citalopram group. Duncko and colleagues examined the effects of SSRIs in an all-male sample also. Thirty-one healthy men were randomised to receive either tianeptine (37.5mg/day), citalopram (20mg/day), or placebo for six days. Following this, the men underwent a stress protocol comprised of a short intelligence test and the Stroop colour interference test. The antidepressant drugs brought about no changes in cortisol stress reactivity compared to placebo. However, after seven days administration, antidepressant treatment in the same male sample brought about enhanced ACTH responses to insulin-induced hypoglycaemia compared to placebo (Jezová & Duncko, 2002). Cortisol remained unaffected.

186

Kotlyar and colleagues have investigated the effects of SSRIs on cortisol stress reactivity in a mixed healthy sample. Using a crossover design, 62 men and women received one month treatment with paroxetine (10-20mg/day) and placebo (Kotlyar et al., 2013). Following treatment participants underwent a modified version of the TSST. Paroxetine brought about a significant overall increase in cortisol levels. However, there was no significant difference in the cortisol stress response between the conditions. Cortisol stress reactivity in depressed patients undergoing treatment with SSRIs has also been examined. Patients with major depression who had been treated with bupropion (200-450mg/day, n=17), or paroxetine (10-50mg/day, n=17) for at least two months were compared to 15 non-depressed controls (Straneva-Meuse, Light, Allen, Golding, & Girdler, 2004). All participants underwent a modified version of the TSST. Those taking bupropion and paroxetine had blunted cortisol stress reactivity compared to healthy controls. 6.3.3 Summary Only one study to date has assessed the effects of acute SSRI treatment on cortisol responses to psychosocial stress (Garcia-Leal et al., 2010). The pharmacological effects of SSRIs are known to be delayed (Frazer & Benmansour, 2002) meaning that longerterm administration may be required to see the effects of SSRIs on cortisol stress reactivity. However, the results from these longer-term studies have been mixed. As with the longer-term beta-blocker studies, SSRIs only seem to enhance stress-related neuroendocrine activity in all-male samples (Jezová & Duncko, 2002; Ljung et al., 2001). Healthy and depressed samples report null findings or blunted cortisol reactivity respectively (Kotlyar et al., 2013; Straneva-Meuse et al., 2004). To date, there have been too few studies carried out on the effects of SSRIs on the cortisol stress response, and

187

amongst the studies that do exist there has been much variability in terms of sample characteristics (e.g. healthy versus depressed versus abdominally obese), treatment duration, and stress protocol used. The evidence suggests that SSRIs affect basal/diurnal cortisol secretion, and that they likely also affect secretion during times of stress. More work is needed to clarify these effects. 6.4 Cortisol stress reactivity: Aims and hypotheses The aim of this study was to examine the effects of seven-day administration of betablockers and SSRIs on cortisol stress reactivity in the laboratory using data from the Stress Pathways Study. As discussed in Chapter 2, chronic stress and depression are associated with changes in the cortisol stress response. Altered cortisol stress reactivity has also been associated with cardiovascular risk factors, and has been seen in CHD patients. Looking at how beta-blockers and SSRIs might alter cortisol secretion after acute laboratory stress in healthy volunteers may provide information about the biological systems involved in stress-related HPA axis dysregulation and may also identify potential therapeutic interventions. Beta-blockers Based on results from studies outlined above, I hypothesise that seven-day treatment with propranolol will bring about increased cortisol stress reactivity in the laboratory. SSRIs Based on results from studies outlined above in healthy volunteers free from depression, I hypothesise that seven-day treatment with escitalopram will bring about increases in cortisol stress reactivity in the laboratory.

188

Moreover, in studies examining longer-term administration of beta-blockers and SSRIs on cortisol stress reactivity, significant enhancement of cortisol secretion has only been reported in all-male samples. Therefore, I will also examine how sex influences the effects of the study medications. 6.5 Literature review: Beta-blockers and corticosteroid receptor sensitivity As mentioned previously, studies using beta-blockers have provided evidence for the notion that suppression of the SNS brings about increased HPA axis activity. It is possible that beta-blockade may modulate HPA axis activity via the corticosteroid receptors. For example, epinephrine and norepinephrine have been shown to affect GR transactivation, and GR binding to GREs within the cell nucleus (Schmidt, Holsboer, & Spengler, 2001). However, there is a dearth of research assessing the effects of beta-blockade on corticosteroid receptor function. To date, one study has assessed the effects of drugs commonly used to treat CHD on GR protein levels, and one of the drugs included was the beta-blocker metoprolol. Measuring GR protein levels gives an indication of GR gene function. Eighty hospitalised CHD patients were enrolled in the study. Twenty of these patients received metoprolol (50mg/day) and GR protein levels in lymphocytes were measured before and one month after administration of the drug (Ji, Guo, Yan, Li, & Lu, 2010). Results indicated that those taking metoprolol had increased GR protein levels following one month of treatment, compared with baseline levels. This result provides support for modulation of the corticosteroid receptors by beta-blockade. However, much more work is needed, particularly in healthy volunteers where the drug effects cannot be ascribed to symptom remission. Examining both the effects of beta-blockade on basal corticosteroid receptor sensitivity and on how the receptors respond to acute stress would provide information on how suppression of the SNS brings about increases in HPA axis function. 189

6.6 Literature review: SSRIs and corticosteroid receptor sensitivity As mentioned previously, depression is known to be characterised by dysregulation of the HPA axis. One explanation of this dysregulation is an imbalance in both GR and MR sensitivity. There is evidence to suggest that SSRIs directly modulate corticosteroid receptor sensitivity. This may be one of the mechanisms through which SSRIs serve to ‘normalise’ HPA axis activity (Anacker, Zunszain, Carvalho, & Pariante, 2011). The effects of SSRIs on both corticosteroid receptor function and sensitivity have been examined in both murine and human studies. 6.6.1 Murine studies Pariante and colleagues examined the effects of 24-hour co-incubation of LMCAT murine cells with dexamethasone and the SSRIs paroxetine, citalopram, and fluoxetine on GR function (Pariante et al., 2001; Pariante, Kim, Makoff, & Kerwin, 2003). GR function was measured by looking at rates of GR-mediated gene transcription. Citalopram, paroxetine, and fluoxetine, were all found to enhance GR function in this cell line. They also found that SSRIs increased GR function by inhibiting the LMCAT cell membrane steroid transporter (a protein, like p-glycoprotein, that expels glucocorticoids from cells). This idea was later disproven when Mason and colleagues showed that the effects of the tricyclic antidepressant desipramine on glucocorticoid accumulation did not differ between wild-type and p-glycoprotein knockout mice (Mason, Thomas, Lightman, & Pariante, 2011). Lai and colleagues assessed the effects of four-day incubation with fluoxetine on GR and MR mRNA expression in rat hippocampal cells (Lai et al., 2003). In line with Pariante’s findings, fluoxetine significantly increased GR mRNA expression. However, MR mRNA expression was unaffected. The authors suggest this shows that GR and MR are 190

differentially regulated by short-term exposure to increased serotonin levels. Interestingly, nine-day incubation with fluoxetine brought about a decrease in MR mRNA expression in rat hippocampal cells (Yau, Noble, Hibberd, & Seckl, 2001). However, this differential regulation in GR and MR function seems to even out following 14-day treatment with SSRIs. In rat hippocampal cells, 14-day incubation with paroxetine brought about increases in GR mRNA expression (Okugawa et al., 1999), and incubation with citalopram for the same time period also brought about increases in MR mRNA expression (Seckl & Fink, 1992). This indicates that longer-term incubation with SSRIs brings about increased GR and MR function. 6.6.2 Human in vitro studies The effects of fluoxetine on GR function has been measured in the lymphocytes of healthy volunteers (Okuyama-Tamura, Mikuni, & Kojima, 2003). In this study GR function was measured by looking at the rate of translocation of the GR into the cell nuclei. Following one hour incubation, fluoxetine induced rapid translocation of the GR into the cell nuclei, meaning this SSRI enhanced GR function. Carvalho and colleagues investigated the effects of a number of different types of antidepressants on GR sensitivity in whole blood drawn from healthy volunteers (Carvalho, Garner, Dew, Fazakerley, & Pariante, 2010). GR sensitivity was measured using dexamethasone inhibition of LPS-stimulated IL-6 production in whole blood. Whole blood was co-incubated for 24 hours with dexamethasone and two different tricyclic antidepressants, one serotonin and norepinephrine reuptake inhibitor (SNRI) and two SSRIs (sertraline and paroxetine). The results indicated that all the antidepressant types brought about reduced GR sensitivity. This finding is in disagreement with previous research which indicates that antidepressants seem to increase GR function in murine and human in vitro studies.

191

6.6.3 Human in vivo studies As mentioned in Section 2.9.2 of this thesis, the DST is the most widely used method to measure GR-mediated negative feedback of the HPA axis in humans in vivo (Rohleder, Wolf, & Kirschbaum, 2003). The dexamethasone/CRH (dex/CRH) test is another version of the DST which is said to be more specific, and have more utility when it comes to diagnosing mood disorders (Watson, Gallagher, Smith, Ferrier, & Young, 2006). The dex/CRH test is essentially a DST followed by a CRH infusion which is supposed to induce the release of ACTH from the pituitary. Like the DST, the dex/CRH test is considered an indirect way to measure corticosteroid receptor sensitivity as dexamethasone administration will modulate the HPA axis via interaction with the corticosteroid receptors. Some argue that the dex/CRH test can detect subtle changes in HPA axis function that the DST cannot (Watson et al., 2006). However, others claim that the dex/CRH test simply measures GR-mediated negative feedback of the HPA axis at both the level of the adrenal and the pituitary glands (Pariante & Miller, 2001). Studies that have investigated the effects of antidepressants on glucocorticoid sensitivity in vivo have almost exclusively used the dex/CRH test. The majority of studies have been carried out in depressed patients. Nickel and colleagues provided six weeks treatment with paroxetine to 22 depressed men and women (Nickel et al., 2003). These patients underwent the dex/CRH test at baseline and at the end of the treatment period. Paroxetine administration resulted in decreases in ACTH and cortisol levels following the dex/CRH tests indicating SSRI-induced increases in GR sensitivity. In a similar study, 20 depressed patients underwent the dex/CRH test following one week of receiving placebo, and after two, four, and 16 weeks of receiving treatment with citalopram (40mg/day) (Nikisch et al., 2005). There was a time-dependent reduction in the ACTH and cortisol responses to the test over the 16-week treatment period indicating an increase in corticosteroid receptor 192

sensitivity. Moreover, the magnitude of the decrease in cortisol responsivity (increase in receptor sensitivity) at four weeks was significantly associated with a reduction in depressive symptoms at 16 weeks. Bschor and colleagues gave 30 patients with depression four weeks SSRI therapy with citalopram (20-40mg/day) (Bschor et al., 2012). The patients underwent the dex/CRH test before and after treatment. Citalopram reduced the amount of ACTH released following the dex/CRH test indicating increased GR sensitivity. Cortisol levels remained unaffected. This implies that citalopram effects took place at the pituitary, but not the adrenal, level of the HPA axis. Reductions in ACTH and cortisol responses to the dex/CRH test have also been observed in 30 female patients with borderline personality disorder receiving 12-week treatment with fluvoxamine (150mg/day) (Rinne et al., 2003). Interestingly, those who had a history of sustained childhood abuse showed the strongest reduction in responses, and they also had the lowest GR sensitivity at baseline. This indicates that SSRI treatment increased GR sensitivity in these patients, particularly in those who had experienced chronic stress in early life. However, some studies have reported contradictory results in depressed patients. In a recent study, 28 patients with major depression received five weeks treatment with escitalopram (10mg/day) (Sarubin, Nothdurfter, Schmotz, et al., 2014). The dex/CRH test was carried out at baseline, and after one and five weeks of treatment with the SSRI. Interestingly, escitalopram led to an increase in cortisol responses to the dex/CRH test after week one, whereas levels of suppression at baseline and five weeks were comparable. What this implies is that treatment with escitalopram brought about a transient decrease in GR sensitivity, but overall had no significant long-term effect. An observational study has also reported decreased GR sensitivity in SSRI users. Manthey and colleagues examined cross-sectional associations between SSRI use and responses to 193

the DST in 1526 patients from the Netherlands Study of Depression and Anxiety (Manthey et al., 2011). Compared to non-users (n=1068), those who used SSRIs (n=309) had decreased cortisol suppression after dexamethasone ingestion. This implies that they had reduced GR sensitivity. The authors controlled for a number of relevant factors including duration of SSRI use, and severity of depression. They posit that treatment resistance, a factor they did not consider in their analysis, may explain their incongruous result. Treatment resistance has been associated with impaired responses to glucocorticoid suppression tests (Juruena et al., 2009). The effects of SSRIs on glucocorticoid sensitivity have also been examined using healthy samples. Carpenter and colleagues administered six weeks treatment with either sertraline (100mg/day) or placebo to 22 healthy men and women (Carpenter et al., 2011). Participants underwent the dex/CRH test at baseline and following the treatment period. The results showed that those who received sertraline had increased cortisol levels following the dex/CRH test compared with placebo. This implies that SSRI treatment in healthy people led to a decrease in GR sensitivity, leading to impaired feedback inhibition of the HPA axis. Pariante and colleagues examined the effects of shorter-term administration of SSRIs in healthy volunteers (Pariante et al., 2004). Eight healthy men and women were given four days administration of citalopram (20mg/day). Participants underwent the PST at baseline and after the four days treatment. Citalopram increased cortisol suppression by prednisolone indicating that this SSRI brought about increased corticosteroid receptor sensitivity (prednisolone binds to both the GR and MR). However, this increase in suppression was only observed in the morning, and not in the early or late afternoon, implying that the diurnal rhythm of HPA axis activity (regulated by the MR) may be an influencing factor here.

194

6.6.4 Summary In murine samples it appears that SSRIs bring about increases in corticosteroid receptor function. Similarly, in vitro examination of GR function in human lymphocytes reveals that SSRIs increase the rate of translocation of the GR into the cell nuclei. However, GR sensitivity, measured using in vitro glucocorticoid sensitivity assays, appears to be decreased in human whole blood incubated with antidepressants. Directly measuring GR function (translocation of receptors into cell nuclei) is different to assessing sensitivity using glucocorticoid sensitivity assays as they provide only a proxy measure of biological receptor ‘function’. This may be a reason for the discrepancy in results. Within depressed patients, the research suggests that SSRI treatment does bring about increases in GR receptor sensitivity as measured by the dex/CRH test. In one case the magnitude of the increase in sensitivity was associated with symptom improvement. These results are in support of the notion that SSRIs may help to ameliorate symptoms of depression by ‘normalising’ dysregulated HPA axis function. Two studies reported decreases in receptor sensitivity in depressed patients (Manthey et al., 2011; Sarubin, Nothdurfter, Schmotz, et al., 2014). However, Manthey and colleague’s study did not adopt an experimental design meaning a number of factors could not be controlled for. In Sarubin and colleague’s study, the participants were also randomised to undergo a yoga intervention (Sarubin, Nothdurfter, Schüle, et al., 2014). Yoga is known to affect cortisol levels (Field, 2011), and this factor was not adjusted for in this study meaning the yoga intervention could have influenced how the SSRIs interacted with the HPA axis. To date, only two studies have assessed the effect of SSRIs on corticosteroid receptor sensitivity in healthy people, eliciting mixed results. Long-term treatment was found to decrease GR sensitivity whereas short-term treatment was found to increase both GR and

195

MR sensitivity in the morning only. Due to the difference in treatment durations, and the use of different suppression tests (dex/CRH versus PST), it is difficult to draw conclusions from the results of these studies. More work is needed examining the effects of SSRIs on corticosteroid receptor function in healthy individuals. Additionally, examining the effects of SSRIs on how corticosteroid receptors sensitivity changes in response to acute stress may shed light on how SSRIs affect cortisol stress reactivity. 6.7 Corticosteroid receptor sensitivity: Aims and hypotheses The aim of this study was to assess the effects of seven-day administration of betablockers and SSRIs on corticosteroid receptor sensitivity both before and after acute psychosocial stress in the laboratory using data from the Stress Pathways Study. As outlined in Chapter 2, chronic stress and depression are associated with alterations in corticosteroid receptor sensitivity. This implies that stress-related HPA axis dysregulation is brought about via diminished sensitivity of these receptors. Examining how betablockers and SSRIs might alter pre-stress baseline receptor sensitivity, and how these drugs affect receptor sensitivity responses to acute stress, may provide more information about how stress-related HPA axis dysregulation comes about. Placebo Acute receptor reactivity: Earlier in this thesis I outlined the both human and animal studies carried out to date examining the effects of acute psychosocial stress in the laboratory on corticosteroid receptor sensitivity (See Chapter 2, Table 2.2). The results of these studies suggest that the effects of acute stress on GR sensitivity vary according to sex, age, BMI, and health status. Overall, there is a lack of studies investigating the effects of acute stress on GR and MR sensitivity, and the observed effects of covariates are yet to be replicated. Due to the variability of results from these studies, and the lack of studies 196

examining the effects of acute stress on the MR, in this study I will examine the effects of acute psychosocial stress on both GR and MR sensitivity in unmedicated healthy volunteers who have received placebo. Based on the findings of human and murine studies outlined in Chapter 2 (Section 2.10.2) and based on the findings from previous work carried out by our group (Carvalho et al., 2015) I hypothesise that acute stress will lead to a decrease in corticosteroid receptor sensitivity in young unmedicated healthy volunteers. Beta-blockers Baseline receptor sensitivity: As beta-blockade induced increases in GR protein levels in CHD patients (Ji et al., 2010), this suggests that beta-blockers increase GR sensitivity. However, the effects of beta-blockers on GR sensitivity directly are not known. To date there have been no studies assessing the effects of beta-blockers on MR sensitivity. I hypothesise that baseline GR and MR sensitivity will be increased in healthy volunteers receiving beta-blockers compared with placebo. Acute stress receptor sensitivity: In Section 6.4 I hypothesise that seven-day treatment with propranolol will bring about increased cortisol stress reactivity in the laboratory. I therefore hypothesise that seven-day treatment with propranolol will bring about enhanced changes in GR and MR sensitivity in response to acute stress compared with placebo. SSRIs Baseline receptor sensitivity: Although results from healthy volunteers have been mixed, the results from murine models and depressed patients seem to suggest that SSRI administration brings about increased GR sensitivity. The results regarding the MR are a

197

little more mixed, but the results of Pariante and colleague’s (2004) study suggest that short-term SSRI administration increases both GR and MR sensitivity, albeit in the morning. Therefore, I hypothesise that seven-day treatment with escitalopram will bring about increases in baseline GR and MR sensitivity compared with placebo. Acute stress receptor sensitivity: In Section 6.4 I hypothesise that seven-day treatment with escitalopram will bring about increased cortisol stress reactivity in the laboratory. Therefore, I hypothesise that seven-day treatment with escitalopram will bring about enhanced changes in GR and MR sensitivity in response to acute stress compared with placebo. It should be noted that I had no hypotheses concerning differences between propranolol and escitalopram. The study was analysed as two parallel comparisons with placebo, rather than contrasting the two active medication conditions. Sex Sex has not been considered in the studies assessing the effects of beta-blockers and SSRIs on corticosteroid receptor functions in humans. However, as I am examining the effects of how sex influences the effects of the study medications on cortisol stress reactivity, I will also explore how sex influences the medication effects on baseline and stress-related corticosteroid receptor sensitivity. Sex differences in how the corticosteroid receptors respond to acute stress have been previously reported (Rohleder et al., 2001). 6.8 Biological measures Data from the Stress Pathways Study were used to test the hypotheses of this study. To reiterate, participants were randomised to receive seven-day administration of either propranolol (80mg/day), escitalopram (10mg/day), or placebo. Following this, they 198

underwent acute psychosocial stress testing in the laboratory. Saliva and blood samples were taken throughout the session for the measurement of cortisol and corticosteroid receptor sensitivity respectively. A detailed account of the methodology is provided in Chapter 4 of this thesis. 6.8.1 Calculation of cortisol parameters in the laboratory A detailed description of the cortisol sampling procedure is provided in Chapter 4, Section 4.4.4 and Section 4.4.7. To reiterate, during the laboratory stress testing session, one saliva sample was taken prior to the stress protocol to allow measurement of baseline cortisol levels (25 minutes after cannulation). A sample was then taken immediately poststress, and at 10, 20, 45, and 75 minutes post-stress to measure cortisol stress reactivity. Cortisol values for each time-point were calculated as well as the overall cortisol AUC in the laboratory (post-stress – 75 minute sample). The cortisol AUC was calculated with respect to ground (Pruessner et al., 2003). Although 91 participants provided complete samples during the laboratory session, not all time-points were included for each participant. Participants were excluded from the analysis if any cortisol value in the laboratory exceeded 50 nmol/L. Therefore, at baseline five values were removed, poststress two values were removed, at 10m post-stress three values were removed, at 20m post-stress and 45m post-stress one value was removed, and at 75m post-stress four values were removed. Cortisol AUC was calculated only for those who provided six usable saliva samples. Therefore, after the removal of outliers from the sample, cortisol AUC was calculated for 85 participants. Different sample sizes for each cortisol time-point are detailed in Table 6.4. Counter to expectations, cortisol levels decreased following the acute stress protocol in every experimental condition (see Figure 6.6, Section 6.10.4). The possible reasons for

199

this will be discussed in Section 6.11.6. Therefore, I categorised the participants according to whether or not they responded to the stress protocol. In accordance with Hamer et al. (2010), participants were considered responders if an increase of ≥1 nmol/L cortisol was detected immediately after the stress protocol, 10 minutes, or 20 minutes post-stress relative to baseline. The number of responders in each medication group is provided in Table 6.4. 6.8.2 Corticosteroid receptor sensitivity Reagents RPMI 1640 medium (Sigma), 500ml, sterile, R8758; foetal calf serum (Gibco 10270); penicillin/streptomycin (Sigma), 500ml, sterile, P4458; LPS (Sigma), 10mg, L2630; dexamethasone (Sigma), D4902; prednisolone (Sigma), P-6004. Protocol Corticosteroid receptor sensitivity was measured using an in vitro glucocorticoid sensitivity assay (see Figure 6.1). Sensitivity was assessed by measuring dexamethasone (GR) and prednisolone (MR) suppression of LPS-induced IL-6 production in whole blood. Whole blood was diluted ten-fold using RPMI 1640 medium supplemented with 10% foetal calf serum, 100IU/ml penicillin, and 100mg/ml streptomycin. LPS and either dexamethasone or prednisolone were added into each well of two 48-well FALCON cell culture plates. The following concentrations of dexamethasone and prednisolone were used: 0M, 5.4 x 10-6M, 1.8 x 10-6M, 5.4 x 10-7M, 1.8 x 10-7M, and 5.4 x 10-8M. Subsequently, 540ml of diluted blood was added to each well. Samples were incubated for 24 hours in a humidified atmosphere containing 5% CO2. After incubation, plates were centrifuged (1000 x g, 4°C, 10mins) and the cell culture supernatant was carefully

200

collected. The samples were then stored at -80°C before being analysed for IL-6 production. IL-6 production analysis was carried out using a commercially available Luminex technology kit for IL-6 from Bio-RAD®. The inter- and intra-assay coefficient of variation (CV) for IL-6 analysis was 13.3% and 7% respectively, and the detection limit was 2.6 pg/ml. Dexamethasone suppression of IL-6 production was assessed using the following concentrations: 0M, 5.4 x 10-6M, 1.8 x 10-6M, 5.4 x 10-7M, 1.8 x 10-7M, and 5.4 x 10-8M dexamethasone. Prednisolone suppression of IL-6 production was assessed using the following concentrations: 0M, 5.4 x 10-6M, 1.8 x 10-6M, 5.4 x 10-7M, and 1.8 x 10-7M prednisolone. IL-6 suppression by 5.4 x 10-8M prednisolone was not assessed as prednisolone had a higher IC50 and 5.4 x 10-8M of prednisolone is not associated with any biological function. The glucocorticoid sensitivity assay 1. Whole blood was diluted with RPMI, foetal calf serum, and penicillin-streptomycin 2. LPS was added all wells of a 48 well plate. Either dexamethasone (GR) or prednisolone (MR) were added in varying concentrations to 40 wells. 3. Whole blood from all four timepoints was then added to each well (in duplicate) 4. The plate was incubated for 24 hours 5. Supernatant was collected and analysed for IL-6 Figure 6.1. The glucocorticoid sensitivity assay

201

Glucocorticoid suppression of IL-6 was calculated by considering LPS-stimulated levels of IL-6 in the absence of either dexamethasone or prednisolone as 100%. Percentage inhibition of IL-6 by the glucocorticoids was then calculated using the following equation:

(

LPS-induced IL-6 levels in the presence of glucocorticoids × 100) = % 𝑖𝑛ℎ𝑖𝑏𝑖𝑡𝑖𝑜𝑛 LPS-induced IL-6 levels in the absence of glucocorticoids

Percentage inhibition for each concentration of dexamethasone and prednisolone was then entered into GraphPad Prism (GraphPad Software Inc., San Diego, CA, USA) in order to calculate the log inhibitory concentration 50% (IC50) values of the dose-response curve of dexamethasone and prednisolone suppression of IL-6 production at each timepoint. The IC50 is the measure of how effective a substance is at inhibiting a specific biological function. It represents the concentration of substance or drug required to bring about 50% inhibition in vitro. Figure 6.2 provides an example of how the IC50 is calculated. Log IC50 values are inversely proportional to glucocorticoid sensitivity. Higher log IC50 values indicate that more dexamethasone or prednisolone was required to suppress IL-6 production by 50%, and this implies that GR and/or MR sensitivity is decreased. 6.8.3 Cardiovascular measures A detailed description of the cardiovascular data measured in the stress laboratory is provided in Chapter 4, Section 4.4.7. To reiterate, all participants were attached to a Finometer® PRO in order to measure BP, heart rate, and cardiac output continuously during the laboratory session. All cardiovascular measures were averaged into mean readings taken from five-minute intervals. There was a five-minute baseline interval (prestress), as well as two five-minute recovery period intervals (+40-45m, and +70-75m).

202

1400 LPS-stimulated IL-6 levels in the absence of glucocorticoids (100%)

1200

IL-6 ng/ml

1000 50% inhibition of IL-6 production

800

IL-6 pg

600

400

200

IC50 = 3.9 x 10-7M

0 0

5.4 x 10-6 1.8 x 10-6 5.40E-06 5.4 x 10-8 1.8 x 10-7 5.4 x 10-7 1.80E-06 5.40E-08 1.80E-07 5.40E-07 DEX (M)

Figure 6.2. Sample calculation of the IC50 from one participant

Cardiovascular measures during the stress protocol were averaged across tasks. Cardiac index (L/min/m2) was calculated by dividing cardiac output (L/min) by the body surface area (m2). 6.9 Statistical analyses Kolmogorov-Smirnov tests were performed to test for normality of the distribution in measures of cortisol, corticosteroid receptor sensitivity (IC50 values), and cardiovascular measures in the laboratory. These normality tests revealed that all corticosteroid receptor sensitivity and cardiovascular measures were normally distributed (all p values > 0.05). However, all measures of salivary cortisol in the laboratory were not normally distributed. Log transformation (base 10) normalised the distributions.

203

As in Chapter 5, the data were analysed using two parallel statistical analyses: propranolol versus placebo, and escitalopram versus placebo. Two-way ANOVAs and chi-square tests were used to compare the three study medication groups on all demographic characteristics. Where possible, sex was included as a between-person factor alongside experimental condition. Paired t-tests were used to assess differences between subjective stress ratings at rest and following the acute stress protocol in the overall sample. Oneway ANOVAs were used to assess the effects of the study medications on subjective stress ratings at rest, and ANCOVAs were used to assess the effects of the study medications on subjective stress ratings following the stress protocol, adjusting for rest ratings. All biological stress measures in the laboratory were analysed using two separate pairwise analyses; propranolol versus placebo, and escitalopram versus placebo. Repeated measures ANOVAs were used to examine stress-related changes over time in cortisol, corticosteroid receptor sensitivity, and cardiovascular measures. Paired t-tests were used to explore significant within-subject contrasts. Where necessary, differences between the two experimental conditions in biological stress parameters at each time-point were analysed using two-way ANOVAs, with experimental condition and sex being included as the main fixed factors. Logistic regression was used to assess the effects of experimental condition on the cortisol responder category. Where there were significant differences between experimental conditions on any of the demographic characteristics, repeated measures ANCOVAs were run where the demographic variable of interest was included as a covariate. Pearson’s R correlations were used in exploratory analysis to ascertain what factors were associated with pre-stress salivary cortisol levels in the laboratory.

204

The significance level was set to p < 0.05 for all analyses, with precise p values reported for all test results. All statistical analyses were performed using SPSS version 22.0 (SPSS Inc., Chicago, Illinois, USA). 6.10 Results 6.10.1 Participants As mentioned previously, 91 participants provided saliva samples for cortisol measurement in the laboratory. Blood used for the measurement of corticosteroid receptor sensitivity was drawn successfully from 85 participants, and cardiovascular measures were gathered from 90 participants. Participants with at least one of these biological measures (cortisol, receptor sensitivity, or cardiovascular measures) were included in the main sample of this study (n = 91). Of the 91 participants in this main sample, 30 were taking escitalopram, 31 were taking propranolol, and 30 received placebo. Table 6.1 summarises the characteristics of the participants in each experimental condition. The sample had an age range of 18-48 years (M = 22.8, SD = 4.8), were almost two-thirds women (63.7%), and were mostly normal weight (78.0% BMI 0.05). Cardiac index is calculated by multiplying heart rate by stroke volume. As cardiac index was elevated during stress and 75 minutes post-stress in the escitalopram condition this indicates that escitalopram increased stroke volume during stress, seeing as heart rate was not affected.

Cardiac index (L/min/m²)

4.5

*

4 3.5

*

Escitalopram Placebo

3 2.5 2 Baseline

Stress

45 mins

75 mins

Figure 6.5. Mean cardiac index values across the session in the escitalopram (blue line) and placebo (grey line) groups. Error bars represent SEM.

216

Taken together, these cardiovascular results indicate that both propranolol and escitalopram were biologically active in these participants as both medications altered heart rate and cardiac index respectively. 6.10.4 Cortisol stress reactivity Overall sample In the overall sample, there was a significant linear main effect of time on cortisol levels across the testing session (F(1, 82) = 84.3, p < 0.001). Pairwise comparisons revealed that there was a significant difference in cortisol levels between each time-point (p values range from < 0.001 – 0.027). However, looking to the mean values indicated that contrary to expectation cortisol levels steadily decreased from baseline values irrespective of the acute stress protocol (see Figure 6.6).

19

15 13 11 9 7

Stress

Cortisol nmol/L

17

5

Figure 6.6. Mean cortisol values (not log-transformed) at each time-point across the testing session in the overall sample. Error bars represent SEM.

217

Table 6.4. Mean (raw) cortisol values across the laboratory session in each experimental condition (p values from analyses with log transformed cortisol values). Propranolol (n=31)

Escitalopram (n=30)

Placebo (n=30)

Propranolol vs. placebo

Escitalopram vs. placebo

Mean ± SD or N(%)

Mean ± SD or N(%)

Mean ± SD or N(%)

Group p value

Group*sex p value

Group p value

Group*sex p value

Baseline cortisol (nmol/L)(n=86) Mean difference±SE

18.0±12.1 2.6±2.9

17.7±10.9 2.3±2.7

15.4±10.0

0.255

0.763

0.597

0.279

Post-stress cortisol (nmol/L)(n=89) Mean difference±SE

13.5±7.5 -0.4±2.3

14.9±8.3 1.0±2.0

13.9±6.7

0.642

0.322

0.630

0.239

Cortisol 10m post-stress (nmol/L)(n=88) Mean difference±SE

12.4±6.7 -1.5±2.3

12.0±5.6 -1.9±2.1

13.9±9.1

0.845

0.619

0.235

0.433

Cortisol 20m post-stress (nmol/L)(n=90) Mean difference±SE

11.4±4.9 -1.2±2.31

11.1±5.7 -1.5±1.84

12.6±7.0

0.540

0.413

0.140

0.399

Cortisol 45m post-stress (nmol/L)(n=90) Mean difference±SE

9.63±3.52 0.25±1.60

9.27±4.59 -0.11±1.46

9.38±4.31

0.214

0.319

0.642

0.756

Cortisol 75m post-stress (nmol/L)(n=87) Mean difference±SE

8.87±2.96 0.09±1.05

8.68±4.21 -0.10±1.09

8.78±3.95

0.194

0.578

0.990

0.704

Cortisol AUC (nmol/L)(n=85) Mean difference±SE

833.7±342.4 15.5±95.6

774.3±367.1 -43.9±100.5

818.2±384.6

0.798

0.871

0.479

0.293

Cortisol responders (n=86)

4(12.9)

7(23.3)

10(33.3)

0.084†

-

0.359†

-

Cortisol stress reactivity (raw scores)

nmol = nanomole. Mean difference is calculated by subtracting placebo values from each experimental condition. The SE of the mean difference was calculated as the square root of the sum of the squares of the SE for each group. †p value for logistic regression

218

Propranolol versus placebo Repeated measures ANOVA revealed no significant main effect of drug (p = 0.98) or time x drug interaction (p = 0.21) on cortisol stress reactivity. The propranolol and placebo groups did not differ significantly on baseline levels of cortisol, post-stress levels of cortisol, or levels of cortisol at 10, 20, 45, and 75 minutes after stress (all p values > 0.05, see Table 6.4). Additionally, there was no difference between groups on overall cortisol output in the laboratory (AUC) (p = 0.80). There was no interactive effect of sex on cortisol levels across the session (all p values > 0.05). Logistic regression was used to test associations between drug type and whether participants were cortisol responders or non-responders. There was no association between experimental condition and cortisol response (p = 0.08). Escitalopram versus placebo Repeated measures ANOVA revealed a significant time x drug quadratic effect (F(1, 52) = 4.49, p = 0.039) which indicates that the slope of change in cortisol across the testing session differed across experimental conditions (see Figure 6.7). Paired t-tests were used to explore this effect further. In the escitalopram group, baseline cortisol values were significantly higher than values during stress (p = 0.001), and at 10 (p < 0.001), 20 (p

Smile Life

When life gives you a hundred reasons to cry, show life that you have a thousand reasons to smile

Get in touch

© Copyright 2015 - 2024 PDFFOX.COM - All rights reserved.